The AI Toolchain — issue -363, August 18, 2025

Agno v1.7.10 adds GPT-5 support, password-protected PDF ingestion, GitHub pagination, and a Team role parameter.

$ git clone --branch v1.7.10 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:
$ git checkout v1.7.10

from agno.team import Team

research_team = Team(
    name='Research Team',
    role='Gather and synthesize information from web sources to answer factual questions',
    members=[...]
)

• ›Adds role parameter to the Team class for defining a team's purpose and specialization.
• ›Adds password-protected PDF support to PDFKnowledgeBase for ingesting secured documents into knowledge bases.
• ›Supports GPT-5 via the OpenAIResponses class.
• ›Adds pagination with metadata for GitHub Tools.

Agno v1.7.9 adds reranker support in PgVector hybrid search and page-number-aware PDF chunking

$ git clone --branch v1.7.9 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:
$ git checkout v1.7.9

• ›Adds stream_intermediate_steps support when using an output model, enabling streaming of intermediate agent steps alongside structured output.
• ›Adds reranker support to PgVector hybrid search, allowing result re-ranking in vector+keyword search pipelines.
• ›Adds page number handling to PDF Readers, including a flag to control whether pages are split during chunking.

Agno v1.7.8 adds output_model for Agents and Teams, OpenAI service_tier support, Gemini thinking, and Google toolkit enhancements.

$ git clone --branch v1.7.8 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:
$ git checkout v1.7.8

from agno.agent import Agent
from agno.models.openai import OpenAIChat

agent = Agent(
    model=OpenAIChat(id='o3'),
    output_model=OpenAIChat(id='gpt-4o-mini'),
)
agent.print_response('Summarise the quarterly results.')

from agno.models.openai import OpenAIChat

model = OpenAIChat(id='gpt-4o', service_tier='flex')

• ›Adds output_model parameter to Agent and Team classes, letting the final response be generated by a separate model rather than the primary model.
• ›Adds service_tier field to OpenAIChat and OpenAIResponses to enable OpenAI Flex Processing.
• ›Adds Gemini thinking output to responses via the Gemini integration.
• ›Adds custom port support for the Google Sheets Toolkit authentication flow.
• ›Enhances Google Calendar toolkit with unified authentication.

AutoGPT Platform v0.6.21 adds TikTok publishing via Ayrshare, updated Exa websets, and five new GitHub integration blocks.

$ git clone --branch autogpt-platform-beta-v0.6.21 https://github.com/Significant-Gravitas/AutoGPT.git
# already have the repo? check out this version:
$ git checkout autogpt-platform-beta-v0.6.21

• ›Adds five new GitHub Integration blocks, expanding the range of GitHub automation workflows available on the platform.
• ›Enables TikTok support through the Ayrshare integration, allowing agents to publish content to TikTok.
• ›Updates the Exa websets implementation with revised block behavior for web research workflows.

AutoGPT Platform adds GPT-5, Claude Opus 4.1, and new OpenAI open-source models, plus ISO 8601 support in time/date blocks.

$ git clone --branch autogpt-platform-beta-v0.6.20 https://github.com/Significant-Gravitas/AutoGPT.git
# already have the repo? check out this version:
$ git checkout autogpt-platform-beta-v0.6.20

• ›Time/date blocks now support ISO 8601 and custom date/time formats.
• ›Adds GPT-5 models to the platform.
• ›Adds OpenAI's new open-source models to the platform.
• ›Adds Anthropic's Claude Opus 4.1 model to the platform.
• ›Separates the notification service from the scheduler as an independent service.

AutoGPT Platform adds Ayrshare YouTube/Instagram, Firecrawl scraping, new LLM models, and AI-generated agent activity status.

$ git clone --branch autogpt-platform-beta-v0.6.19 https://github.com/Significant-Gravitas/AutoGPT.git
# already have the repo? check out this version:
$ git checkout autogpt-platform-beta-v0.6.19

• ›Adds Firecrawl integration for web scraping and data extraction as a new block/service connection.
• ›Enables Ayrshare YouTube support, allowing agents to publish or interact with YouTube via Ayrshare.
• ›Enables Ayrshare Instagram support, allowing agents to publish or interact with Instagram via Ayrshare.
• ›Adds new LLM models to the platform's model selection.
• ›Adds AI-generated activity status for agent executions, surfacing real-time descriptive status during runs.

CrewAI 0.157.0 adds a crewai config CLI command group, Okta device authorization, LangDB integration, and initial tracing capabilities.

$ git clone --branch 0.157.0 https://github.com/crewAIInc/crewAI.git
# already have the repo? check out this version:
$ git checkout 0.157.0

$ crewai config

• ›Adds crewai config CLI command group for managing CLI configuration.
• ›Supports device authorization with Okta for authenticated flows.
• ›Adds LangDB integration for observability and query tracing.
• ›Introduces initial tracing capabilities for crew and flow execution.
• ›Enables persisting Flow state with BaseModel entries.

langchain-openai 0.3.29 adds minimal/verbosity response control, custom tools support, and prompt_cache_key parameter.

$ git clone --branch langchain-openai==0.3.29 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-openai==0.3.29

• ›Adds minimal and verbosity parameters to control response detail level in OpenAI chat completions.
• ›Adds custom tools support, enabling users to pass custom tool definitions to OpenAI models.
• ›Adds prompt_cache_key parameter support for controlling prompt caching behavior.
• ›Adds max_retries parameter to ChatOpenAI for handling 503 capacity errors.

langchain-core 0.3.73 zeros out token costs for cache hits.

$ git clone --branch langchain-core==0.3.73 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-core==0.3.73

• ›Token costs are now zeroed out for cache hits, preventing inflated cost tracking when cached responses are returned.

LangChain 0.4.0.dev0 introduces standard outputs as a new capability.

$ git clone --branch langchain==0.4.0.dev0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain==0.4.0.dev0

• ›Adds standard outputs support to LangChain.

langchain-openai 0.4.0.dev0 adds standard structured outputs support to ChatOpenAI.

$ git clone --branch langchain-openai==0.4.0.dev0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-openai==0.4.0.dev0

• ›Adds standard outputs support (structured output schema handling) to the OpenAI integration.

langchain-core 0.4.0.dev0 introduces standard outputs for structured LLM responses.

$ git clone --branch langchain-core==0.4.0.dev0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-core==0.4.0.dev0

• ›Adds standard outputs support, providing structured response formats for LLM outputs.

langchain-groq 0.3.7 loosens reasoning_effort restrictions and adds OpenAI-OSS model support.

$ git clone --branch langchain-groq==0.3.7 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:
$ git checkout langchain-groq==0.3.7

• ›Loosens restrictions on reasoning_effort and injects effort value into response metadata for Groq calls.
• ›Adds support for OpenAI-OSS models via the Groq integration.

Letta 0.11.1 adds Claude Opus 4.1 and GPT-5 support, a new minimal reasoning effort option, and paginated file grep.

$ git clone --branch 0.11.1 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.11.1

from letta import LLMConfig

config = LLMConfig(
    model="gpt-5",
    reasoning_effort="minimal"
)

• ›Adds minimal as a valid value for the reasoning_effort parameter in LLMConfig, giving finer control over model reasoning cost.
• ›Adds support for Claude Opus 4.1 and GPT-5 models.
• ›Makes the built-in grep tool for files paginated, enabling traversal of large file search results.

Letta 0.11.0 adds SigNoz OTEL trace export and a filesystem demo with file upload and streaming.

$ git clone --branch 0.11.0 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.11.0

$ docker run \
  -v ~/.letta/.persist/pgdata:/var/lib/postgresql/data \
  -p 8283:8283 \
  -e SIGNOZ_ENDPOINT=${SIGNOZ_ENDPOINT} \
  -e SIGNOZ_INGESTION_KEY=${SIGNOZ_INGESTION_KEY} \
  -e LETTA_OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317 \
  letta/letta:latest

• ›Adds SigNoz integration for exporting OpenTelemetry traces by setting SIGNOZ_ENDPOINT, SIGNOZ_INGESTION_KEY, and LETTA_OTEL_EXPORTER_OTLP_ENDPOINT environment variables.
• ›Adds filesystem demo with file upload and streaming support.
• ›Jinja template rendering is now offloaded to the thread pool, reducing CPU-bound blocking of the async event loop.

Letta 0.10.0 adds LettaPing keepalives for long streaming connections, MCP OAuth support, and a new default agent architecture.

$ git clone --branch 0.10.0 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:
$ git checkout 0.10.0

• ›Adds LettaPing message type sent every 90 seconds on streaming endpoints to prevent connection termination during long-running tool calls.
• ›Adds not_indexable property to agents, allowing agents to be excluded from indexing.
• ›Defaults to the new memgpt_v2_agent base architecture; archival memory tools are no longer added by default but can be added explicitly.
• ›Adds OAuth support for MCP providers, enabling integrations with services such as Linear and GitHub.
• ›Adds LMStudio support for Qwen and Llama models with manual token counting for streaming.

LlamaIndex v0.13.1 adds Heroku LLM integration, Bedrock AgentCore toolspecs, voyage context embeddings, BM25 metadata filtering, and more.

$ git clone --branch v0.13.1 https://github.com/run-llama/llama_index.git
# already have the repo? check out this version:
$ git checkout v0.13.1

• ›Adds metadata filtering support to BM25Retriever (via llama-index-retrievers-bm25 0.6.2), enabling filtered keyword retrieval alongside vector stores.
• ›Adds llama-index-llms-heroku 0.1.0, a new LLM integration for Heroku-hosted models.
• ›Adds llama-index-tools-aws-bedrock-agentcore 0.1.0 with toolspecs for Bedrock AgentCore browser and code interpreter.
• ›Adds voyage context embeddings support to llama-index-embeddings-voyageai 0.4.1.
• ›Adds Anthropic citations to non-beta (GA) support in llama-index-llms-anthropic 0.8.2.

OpenAI Agents SDK v0.2.7 adds reasoning.effort and verbosity params to ModelSettings plus a Realtime handoff prompt prefix.

$ git clone --branch v0.2.7 https://github.com/openai/openai-agents-python.git
# already have the repo? check out this version:
$ git checkout v0.2.7

• ›Adds reasoning.effort='minimal' and verbosity parameters to ModelSettings for fine-grained control over model reasoning behaviour.
• ›Adds a handoff prompt prefix for Realtime agents, improving context handoff in real-time sessions.
• ›Adds runtime validation for Agent constructor arguments, catching misconfiguration at instantiation time.

OpenAI Agents SDK v0.2.6 adds output guardrails for realtime agents and logprobs to ModelSettings.

$ git clone --branch v0.2.6 https://github.com/openai/openai-agents-python.git
# already have the repo? check out this version:
$ git checkout v0.2.6

from agents import Agent, ModelSettings

agent = Agent(
    name='analyzer',
    model='gpt-4o',
    model_settings=ModelSettings(logprobs=True)
)

• ›Adds logprobs field to ModelSettings class, enabling log-probability output from model responses.
• ›Supports agent output guardrails in realtime sessions, bringing parity with non-realtime guardrail enforcement.

OpenAI Agents SDK v0.2.5 adds realtime speed control, agent-update-mid-session, MCP server visualization, and split stream events.

$ git clone --branch v0.2.5 https://github.com/openai/openai-agents-python.git
# already have the repo? check out this version:
$ git checkout v0.2.5

async for event in runner.stream():
    if event.type == 'tool_call_item':
        print('Tool invoked:', event.item)
    elif event.type == 'tool_call_output_item':
        print('Tool output:', event.item)

• ›Adds speed parameter to the realtime API to control the pace of model responses during a session.
• ›Adds the ability to update an agent's configuration during an active realtime session via the new update-agent functionality.
• ›Separates tool_call_item and tool_call_output_item into distinct stream events, giving handlers finer-grained control over tool call lifecycle.
• ›Exports MultiProvider in the public API, making multi-model-provider routing directly importable from the agents module.
• ›Visualization now draws MCP servers in agent graphs, making the full tool topology visible.

PydanticAI v0.7.1 adds GPT-5 models, OpenAI verbosity support, pre-request token counting via Gemini, and a new model inference string.

$ git clone --branch v0.7.1 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v0.7.1

from pydantic_ai import Agent

agent = Agent('openai-responses:gpt-4o')
result = await agent.run('What is the capital of France?')
print(result.output)

• ›Adds UsageLimits.count_tokens_before_request to count tokens using Gemini's count_tokens API before a request is sent, enabling proactive limit enforcement.
• ›Supports the "openai-responses" model inference string for selecting the OpenAI Responses API via string-based model configuration.
• ›Adds support for the OpenAI verbosity parameter in the Responses API.
• ›Adds new OpenAI GPT-5 models to the supported model list.

PydanticAI v0.7.0 adds Temporal workflow support, dynamic toolsets, event stream handlers, and new agent abstractions.

$ git clone --branch v0.7.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v0.7.0

async with agent.run_stream('Analyze logs', event_stream_handler=my_handler) as response:
    async for chunk in response.stream_text():
        print(chunk)

• ›Adds event_stream_handler parameter to agent and run methods for subscribing to agent event streams.
• ›Adds Agent.override(tools=...) to replace or inject tools into an existing agent at runtime.
• ›Adds AbstractAgent and WrapperAgent base classes for building composable agent wrappers.
• ›Enables running Agent inside a Temporal workflow by dispatching model requests, tool calls, and MCP as Temporal activities.
• ›Supports dynamically building toolsets based on run context via the toolset API.

PydanticAI v0.6.2 adds builtin_tools parameter to the Agent class.

$ git clone --branch v0.6.2 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v0.6.2

• ›Adds builtin_tools parameter to Agent, enabling control over which built-in tools are available to an agent.

PydanticAI v0.6.1 adds automatic OpenAI strict mode, Bedrock thinking parts, AWS bearer token support, and new Heroku models.

$ git clone --branch v0.6.1 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v0.6.1

• ›Supports AWS_BEARER_TOKEN_BEDROCK environment variable for authenticating with AWS Bedrock via bearer token.
• ›Makes InlineDefsJsonSchemaTransformer part of the public API, allowing direct use in custom JSON schema transformations.
• ›Automatically enables OpenAI strict mode for output types that are strict-compatible, removing the need for manual configuration.
• ›Sends ThinkingParts back to Anthropic when accessed through AWS Bedrock, enabling extended thinking round-trips.
• ›Adds new Heroku models to the supported model list.

PydanticAI v0.6.0 adds a new Anthropic model and removes a wave of long-deprecated APIs.

$ git clone --branch v0.6.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v0.6.0

• ›Adds a new Anthropic model (removes older deprecated Anthropic models in the same change).

PydanticAI v0.5.0 expands OpenAI strict JSON mode compatibility and adds default values to tool argument schemas.

$ git clone --branch v0.5.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v0.5.0

• ›Enables more BaseModels to use OpenAI strict JSON mode by defaulting additionalProperties=False automatically.
• ›Supports string format, pattern, and related constraints within OpenAI strict JSON mode.
• ›Includes default values in the JSON schema generated for tool arguments.

PydanticAI v0.4.11 adds AG-UI convenience functions and custom thinking-tag support on model profiles.

$ git clone --branch v0.4.11 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v0.4.11

• ›Supports custom thinking tags specified on the model profile, letting callers control how chain-of-thought tokens are surfaced per model.
• ›Adds convenience functions to handle AG-UI requests with request-specific dependencies.

Semantic Kernel .NET 1.62.0 adds ONNX provider/CUDA support, HttpClient injection for Azure OpenAI text-to-image, and A2A SDK integration.

$ git clone --branch dotnet-1.62.0 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout dotnet-1.62.0

• ›Adds HttpClient parameter to AddAzureOpenAITextToImage method, enabling custom HTTP client injection for Azure OpenAI text-to-image calls.
• ›AddOpenAIEmbeddingGenerator now respects HttpClient.BaseAddress for endpoint resolution, enabling proxy and custom endpoint scenarios.
• ›Adds execution provider support to the ONNX connector, including a CUDA sample, enabling GPU-accelerated local inference.
• ›Updates the A2A agent integration to use the latest A2A .NET SDK.
• ›Magentic orchestration now returns the last agent message when limits are reached.

Semantic Kernel Python 1.35.1 adds AzureAI MCP tool streaming, Bedrock model provider param, and plugin encoding support.

$ git clone --branch python-1.35.1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout python-1.35.1

kernel.add_plugin(plugin_name='MyPlugin', parent_directory='./plugins', encoding='latin-1')

service = BedrockChatCompletion(model_id='anthropic.claude-v2', model_provider='anthropic')

• ›Adds tool_call_id as a required parameter for string-based tool messages in ChatHistory, enforcing correct message threading.
• ›Adds ability to specify encoding when adding a plugin via add_plugin.
• ›Adds model_provider parameter for Bedrock model configuration.
• ›Supports AzureAI agent MCP tools for both streaming and non-streaming invocations.
• ›Improved MCP connect flow with additional samples.

browser-use 0.5.8 adds a helper function for retrieving formatted agent traces with metadata.

$ git clone --branch 0.5.8 https://github.com/browser-use/browser-use.git
# already have the repo? check out this version:
$ git checkout 0.5.8

• ›Adds a helper function to retrieve a nicely formatted agent trace with metadata for post-run inspection and debugging.

Cline v3.25.0 adds auto context compaction, Focus Chain deep planning, Requesty provider support, and external 'Add to Cline' triggers.

$ git clone --branch v3.25.0 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout v3.25.0

• ›Adds auto compact context management to automatically handle context window limits during long tasks.
• ›Adds Focus Chain & Deep Planning mode, now enabled by default, for structured multi-step task planning.
• ›Adds a retry button for broken streaming responses, letting users recover without restarting a task.
• ›Allows external hosts to trigger the 'Add to Cline' action via the external host bridge.
• ›Adds Requesty as a supported provider integration.

Cline v3.24.0 adds clickable file links in chat, CLINE_ACTIVE terminal env var, browser argument support, new GPT-5 and Kimi K2 models, and context window display in model info.

$ git clone --branch v3.24.0 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout v3.24.0

• ›Adds clickable file names in chat that jump directly to the editor.
• ›Sets CLINE_ACTIVE environment variable in new terminals spawned by Cline, enabling scripts to detect when running inside Cline.
• ›Enables custom browser arguments via browser settings.
• ›Adds OpenAI GPT-5 Chat (gpt-5-chat-latest) model support.
• ›Adds Kimi K2 Turbo Preview (kimi-k2-turbo-preview) model support.

Cline v3.23.0 adds prompt caching support for SAP AI Core models.

$ git clone --branch v3.23.0 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout v3.23.0

• ›Supports prompt caching for SAP AI Core, reducing latency and token costs on repeated context.

Cline v3.21.0 adds GPT-5 support, Ollama cloud endpoint key, and HostBridge diff diagnostics.

$ git clone --branch v3.21.0 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout v3.21.0

• ›Adds support for GPT-5 as an available model.
• ›Supports Ollama API key for cloud-hosted Ollama endpoints.
• ›Adds HostBridge diff diagnostics for improved code change feedback.
• ›Adds onboarding walkthrough button to surface quick wins for new users.

Cline v3.20.13 adds prompt caching support for Claude Opus 4.1, reducing latency and cost.

$ git clone --branch v3.20.13 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout v3.20.13

• ›Enables prompt caching for the Claude Opus 4.1 model, reducing repeated-context latency and token costs.

Cline v3.20.11 adds Cerebras gpt-oss-120b and Claude Opus 4.1 support across Anthropic and Bedrock providers.

$ git clone --branch v3.20.11 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout v3.20.11

• ›Adds Cerebras gpt-oss-120b as a supported model.
• ›Adds Claude Opus 4.1 support for the claude-code provider.
• ›Adds Claude Opus 4.1 support for Amazon Bedrock.
• ›Adds getCallbackUri to the HostProvider API surface.

Cline v3.20.9 adds Baseten provider, Anthropic Opus 4.1, plan-mode file-edit enforcement, and space-in-filename mention support.

$ git clone --branch v3.20.9 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout v3.20.9

• ›Adds Baseten as a new LLM provider integration.
• ›Adds Anthropic Opus 4.1 model support.
• ›Adds optional enforcement to disallow file edits while in plan mode.
• ›Supports @mentions for filenames that contain spaces.

Cline v3.20.4 adds Grok 4 and updated Cerebras models to the provider lineup.

$ git clone --branch v3.20.4 https://github.com/cline/cline.git
# already have the repo? check out this version:
$ git checkout v3.20.4

• ›Adds Grok 4 to the advanced models list for xAI provider support.
• ›Updates available Cerebras models.

Continue v1.0.20 adds parallel tool calling, LLM stream retries, conversation compaction, watsonx support, and Qwen3 autocomplete.

$ git clone --branch v1.0.20-vscode https://github.com/continuedev/continue.git
# already have the repo? check out this version:
$ git checkout v1.0.20-vscode

context:
  - provider: docs
    params:
      useLocalCrawling: true
      maxDepth: 3

• ›Adds useLocalCrawling and maxDepth settings for indexed documents, restoring control over local doc crawl behavior.
• ›Adds slug field to rule objects in dev data.
• ›Supports parallel tool calling, enabling multiple tool calls to execute simultaneously in a single assistant turn.
• ›Adds automatic LLM streamChat retries to recover from transient stream failures.
• ›Adds conversation compaction with a UI button to compress long chat histories and stay within context limits.

Crush v0.6.2 adds Android/Termux support for on-the-go use.

$ git clone --branch v0.6.2 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.6.2

• ›Adds official Android/Termux release, enabling Crush to run on mobile Android terminals.

Crush v0.6.1 adds configurable MCP timeouts, automatic MCP reconnection, and per-filetype LSP scoping.

$ git clone --branch v0.6.1 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.6.1

{
  "mcp": {
    "example": {
      "type": "http",
      "url": "https://example.org/mcp/",
      "timeout": 30
    }
  }
}

{
  "lsp": {
    "Go": {
      "command": "gopls",
      "filetypes": ["go", "mod", "sum", "work"]
    }
  }
}

• ›Adds configurable per-MCP timeout in config (default 15 seconds) so unreliable MCP servers don't hang Crush indefinitely.
• ›Enables automatic MCP client ping and reconnection when an MCP server becomes unresponsive.
• ›Supports restricting LSP servers to specific file types via a filetypes config key for improved efficiency.
• ›Ships built-in filetype associations for 17 popular language servers (gopls, rust-analyzer, pyright, typescript-language-server, and more), so no manual scoping is needed for common LSPs.

Crush v0.6.0 adds Power Steering, letting you queue messages to the model mid-task without interrupting its tool calls.

$ git clone --branch v0.6.0 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.6.0

• ›Adds Power Steering: queue messages for the model to pick up between tool calls or after task completion, with Esc to clear the queue.

Crush v0.5.0 adds configurable default diff mode (unified or split) via the TUI config.

$ git clone --branch v0.5.0 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.5.0

{
  "options": {
    "tui": {
      "diff_mode": "split"
    }
  }
}

• ›Adds diff_mode config option under options.tui to set the default diff view to unified or split.

Crush v0.4.0 adds in-session Yolo Mode toggle, diff view mouse scrolling, built-in coreutils on Windows, and CRUSH_-prefixed env vars.

$ git clone --branch v0.4.0 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.4.0

$ crush --yolo

• ›Adds in-session Yolo Mode toggle via ctrl+p menu, so you can enable/disable permission bypass without restarting Crush.
• ›Enables mouse scrolling in the diff view.
• ›Ships built-in coreutils (base64, gzip, mktemp, shasum, tar) on Windows — and speeds them up on all platforms by eliminating shell-outs.
• ›Supports CRUSH_-prefixed environment variables (e.g. CRUSH_ANTHROPIC_API_KEY) to scope secrets to Crush without clobbering default env vars.

Crush v0.3.0 adds mouse text selection, drag-and-drop image input, GPT-5 support, and OpenAI reasoning model improvements.

$ git clone --branch v0.3.0 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.3.0

• ›Adds mouse-based text selection in chat messages, including double-click to select and copy.
• ›Supports drag-and-drop image input into the chat from almost any terminal, including Apple Terminal.
• ›Adds out-of-the-box support for GPT-5 as an OpenAI provider.
• ›Supports minimal reasoning mode for OpenAI reasoning models.
• ›Available on Nixpkgs unstable for easier installation.

Crush v0.2.2 adds JetBrains rendering support, Scoop distribution for Windows, and LSP/MCP status indicators in the UI.

$ git clone --branch v0.2.2 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.2.2

$ scoop bucket add charm https://github.com/charmbracelet/scoop-bucket.git && scoop install crush

• ›Adds vastly improved rendering support for JetBrains terminals (JediTerm) used in GoLand, IntelliJ, PyCharm, and similar IDEs.
• ›Adds Scoop package distribution for easy installation on Windows.
• ›Adds LSP and MCP status indicators in the UI showing offline, starting up, online, and error states; full details viewable via crush logs.
• ›Adds Escape key as an additional way to deny prompts in the UI.

Crush v0.2.0 adds .crushignore support and enhanced debug logging with provider response capture.

$ git clone --branch v0.2.0 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.2.0

$ crush --debug && crush logs

• ›Adds .crushignore file support to exclude files from Crush's context without removing them from version control; uses .gitignore syntax and works in the project root and subdirectories.
• ›Extends --debug mode to also log provider request/response details, stored in .crush/logs/crush.log and viewable via crush logs.
• ›Applies .crushignore and .gitignore rules to the built-in grep tool, scoping searches consistently with context exclusions.

Goose v1.2.0 adds parallel sub-recipe execution, LiteLLM provider, recipe retry logic, wakelock, and more.

$ git clone --branch v1.2.0 https://github.com/block/goose.git
# already have the repo? check out this version:
$ git checkout v1.2.0

• ›Adds experimental support for running sub-recipes multiple times in parallel for concurrent task execution.
• ›Adds LiteLLM provider with automatic prompt caching support.
• ›Adds recipe retry logic with configurable success criteria so recipes can self-heal on failure.
• ›Adds wakelock feature to prevent the system from sleeping while Goose is working.
• ›Adds inline Python extension support.

Codex CLI v0.21.0 adds diff previews for apply_patch, output truncation notices, and MCP JSON schema sanitization.

$ git clone --branch rust-v0.21.0 https://github.com/openai/codex.git
# already have the repo? check out this version:
$ git checkout rust-v0.21.0

• ›Shows a diff preview when apply_patch is about to modify files, letting you inspect changes before they land.
• ›Includes an output truncation message in tool call results so the model knows when it received incomplete command output.
• ›Adds JSON schema sanitization for MCP tools to ensure compatibility with internal JsonSchema enums.
• ›Sends a prompt cache key to improve token efficiency through better caching.
• ›Splits multiline commands in the UI for improved readability.

Codex CLI v0.15.0: gpt-5 default, new approval mode, and trust-based onboarding flow

$ git clone --branch rust-v0.15.0 https://github.com/openai/codex.git
# already have the repo? check out this version:
$ git checkout rust-v0.15.0

$ codex --ask-for-approval on-request 'refactor the auth module to use JWT'

$ codex --sandbox workspace-write --ask-for-approval on-request 'add unit tests for src/api.rs'

• ›Sets gpt-5 as the default model for all sessions.
• ›Adds --ask-for-approval on-request mode, letting the model decide when to prompt for user approval — a middle ground between on-failure and never.
• ›Introduces a new onboarding flow that auto-configures --sandbox workspace-write and --ask-for-approval on-request when a folder is marked as trusted, optimized for Git repo workflows.

Codex CLI gets native terminal scrolling, /compact command, experimental planning tool, and version-update notifications.

$ git clone --branch rust-v0.11.0 https://github.com/openai/codex.git
# already have the repo? check out this version:
$ git checkout rust-v0.11.0

$ /compact

• ›New terminal UX appends output to the end of the terminal instead of running fullscreen, enabling native scrollback and copy/paste.
• ›Adds the /compact command for compacting conversation context.
• ›Introduces an experimental planning tool.
• ›CLI now checks for newer versions and notifies you when an update is available.

OpenCode v0.5.5 retains cache when cycling between subagent and parent sessions for faster TUI performance.

$ git clone --branch v0.5.5 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.5.5

• ›Retains cache when cycling between subagent and parent sessions, reducing redundant recomputation in the TUI.

OpenCode v0.5.3 adds configurable scroll speed, message navigation modal, and .zshenv loading with ZDOTDIR support.

$ git clone --branch v0.5.3 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.5.3

# In your opencode config file:
scroll_speed: 5

• ›Adds scroll_speed configuration option to control TUI scroll behavior.
• ›New message navigation modal lets you jump to and restore specific messages in a conversation.
• ›Shell integration now loads .zshenv and respects ZDOTDIR when present.

OpenCode v0.5.2 adds Vue LSP support, subagent navigation, reverse model cycling, and new TUI/API capabilities.

$ git clone --branch v0.5.2 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.5.2

• ›Adds Vue LSP support and enables ESLint for .vue files.
• ›Supports cycling recent models in reverse order in the TUI.
• ›Renders tool metadata after permission rejection so users can inspect what was blocked.
• ›New /tui/show-toast API endpoint for displaying toast notifications.
• ›New API routes to get a session and its session children.

OpenCode v0.4.44 converts deleted attachments to text, preserving context in the conversation.

$ git clone --branch v0.4.44 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.44

• ›Converts attachments to text when deleted, keeping their content available in the conversation context.

OpenCode v0.4.43 adds bash command execution and attachment-to-text conversion.

$ git clone --branch v0.4.43 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.43

$ !ls -la

• ›Adds bash command execution via ! prefix, letting you run shell commands directly from the OpenCode CLI.
• ›Supports converting attachments to text when deleted, preserving content as inline text rather than losing it.

OpenCode v0.4.40 adds session renaming in the TUI and a revamped agent modal UI.

$ git clone --branch v0.4.40 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.40

• ›New session rename functionality available directly from the TUI modal.
• ›Revamped UI/UX for the agent modal with improved layout and experience.
• ›Adds OpenAI prompt cache key support.

OpenCode v0.4.27 adds wildcard matching for tool names in config

$ git clone --branch v0.4.27 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.27

• ›Supports wildcard pattern matching for tool names in configuration, enabling bulk enable/disable of tool groups without listing each name explicitly.

OpenCode v0.4.20 adds -c and -s args to the tui command, matching the run command pattern.

$ git clone --branch v0.4.20 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.20

$ opencode tui -c "review my last commit for security issues" -s <session-id>

• ›Adds -c and -s flags to the tui command, aligning its interface with the run command pattern.

OpenCode v0.4.19 adds a toggle to show or hide tool call details in the UI.

$ git clone --branch v0.4.19 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.19

• ›Adds toggle to show or hide tool details visibility in the interface.

OpenCode v0.4.17 adds wrap-around navigation in Models/Tools modals and a placeholder for pending assistant messages.

$ git clone --branch v0.4.17 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.17

• ›Adds wrap-around navigation for list selection in the Models and Tools modals.
• ›Adds a placeholder display on pending assistant messages in the TUI.
• ›Makes the models key optional in config.

OpenCode v0.4.15 adds an agents dialog, F2 model cycling, and biome.jsonc config support.

$ git clone --branch v0.4.15 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.15

$ # While inside opencode TUI, press F2 to step through the 5 most recent models

• ›New agents dialog in the TUI for managing and switching between agents.
• ›Press F2 to cycle through the 5 most recently used models from anywhere in the TUI.
• ›Supports biome.jsonc (JSON with comments) as a Biome config file.

OpenCode v0.4.5 adds ESLint LSP, thinking block rendering, webfetch permissions, and improved attachment pasting.

$ git clone --branch v0.4.5 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.5

• ›Adds ESLint LSP integration for real-time lint feedback within the coding agent.
• ›Renders thinking blocks in both the TUI and share page for supported models.
• ›Adds webfetch permission support, giving users control over agent web-fetch requests.
• ›Improves file attachment pasting workflow.
• ›Displays agent name in the LLM response footer with updated ordering.

OpenCode v0.4.2 adds mode selection when creating agents.

$ git clone --branch v0.4.2 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.2

• ›Supports mode selection at agent creation time, letting you pick the operating mode (e.g., plan or build) upfront.

OpenCode v0.4.0 merges agents and modes into a single unified agent concept with tab-cycling and @-mention spawning.

$ git clone --branch v0.4.0 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.4.0

$ # Press <tab> during an active OpenCode session to cycle to the next configured agent

• ›Unifies agent and mode into a single agent config field — configure one concept instead of two.
• ›Cycle through primary agents with <tab> during a session.
• ›Spawn a subagent inline by @-mentioning it by name.
• ›Agents with a description field are automatically selected by the primary agent based on context.
• ›Re-enables the todo tool for task tracking within sessions.

OpenCode v0.3.132 adds permissions disallow support, improved assistant message visuals, and a config option to disable snapshots.

$ git clone --branch v0.3.132 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.3.132

# In your opencode config file:
snapshot: false

• ›Adds permissions disallow support to restrict what actions the assistant is allowed to perform.
• ›Introduces a snapshot field in config to disable session snapshots.
• ›Improves assistant message visual presentation.

OpenCode v0.3.113 enables Kimi k2 ↔ Claude trajectory handoff for multi-model workflows.

$ git clone --branch v0.3.113 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.3.113

• ›Enables Kimi k2 ⇄ Claude trajectory handoff, allowing sessions to transfer context between Kimi k2 and Claude models.

OpenCode v0.3.112 adds provider search, configurable LSP, configurable top_p, and forced models.dev refresh on login.

$ git clone --branch v0.3.112 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v0.3.112

• ›Adds search filtering in the provider selection UI.
• ›Supports configurable LSP (language server protocol) settings.
• ›Makes top_p model parameter configurable.
• ›Forces models.dev refresh automatically on auth login.

OpenHands 0.52.0 adds GPT-5, Claude Opus 4.1, Qwen-3-Coder-480B, CLI UI launch, and overlay mount support.

$ git clone --branch 0.52.0 https://github.com/All-Hands-AI/OpenHands.git
# already have the repo? check out this version:
$ git checkout 0.52.0

• ›Adds qwen-3-coder-480b to the OpenHands provider model roster.
• ›Adds gpt-5-2025-08-07 to the OpenHands provider model roster.
• ›Adds claude-opus-4-1-20250805 as a supported model.
• ›Adds search API key settings configuration to the CLI.
• ›Supports launching the full OpenHands UI directly from the CLI.

Qwen Code v0.0.6-nightly.1 adds OpenRouter support, Qwen OAuth, systemPromptMappings config, and a telemetry service.

$ git clone --branch v0.0.6-nightly.1 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.6-nightly.1

• ›Adds systemPromptMappings configuration feature to map system prompts per context.
• ›Adds Qwen OAuth integration for authentication flows.
• ›Supports OpenRouter as a model backend provider.
• ›Adds telemetry service for usage tracking.
• ›Makes /init respect the configured context filename, aligning with QWEN.md.

Qwen Code v0.0.6-nightly.0 adds OpenRouter support, Qwen OAuth, systemPromptMappings config, and a telemetry service.

$ git clone --branch v0.0.6-nightly.0 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.6-nightly.0

• ›Adds systemPromptMappings configuration feature to map system prompts per context.
• ›Adds Qwen OAuth integration for authenticated sessions.
• ›Supports OpenRouter as a model provider backend.
• ›Adds telemetry service for usage tracking.
• ›Adds usage statistics logging for the Qwen integration.

Qwen Code v0.0.6 adds usage statistics logging and makes /init respect configured context filenames.

$ git clone --branch v0.0.6 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.6

• ›Adds usage statistics logging for Qwen API integration to track token and request consumption.
• ›Makes /init command respect the configured context filename (e.g., QWEN.md) rather than hardcoding a default.

Qwen Code v0.0.5-nightly.12 adds OpenRouter support, Qwen OAuth, systemPromptMappings config, and usage statistics logging.

$ git clone --branch v0.0.5-nightly.12 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.5-nightly.12

• ›Adds systemPromptMappings configuration feature to map system prompts per context.
• ›Adds Qwen OAuth integration for authenticated sessions.
• ›Supports OpenRouter as a model provider backend.
• ›Adds telemetry service for usage tracking.
• ›Adds usage statistics logging for Qwen integration.

Qwen Code v0.0.5-nightly.11 adds OpenRouter support, Qwen OAuth, systemPromptMappings config, and a telemetry service.

$ git clone --branch v0.0.5-nightly.11 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.5-nightly.11

• ›Adds systemPromptMappings configuration feature to map models to custom system prompts.
• ›Adds OpenRouter as a supported inference provider.
• ›Adds Qwen OAuth integration for authentication.
• ›Adds a telemetry service for usage tracking.
• ›Updates the /bug command to point to the Qwen-Code repository.

Qwen Code v0.0.5-nightly.10 adds OpenRouter support, Qwen OAuth, systemPromptMappings config, and a telemetry service.

$ git clone --branch v0.0.5-nightly.10 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.5-nightly.10

• ›Adds systemPromptMappings configuration feature, enabling per-context system prompt overrides via config.
• ›Adds Qwen OAuth integration for authenticated access to Qwen services.
• ›Adds support for OpenRouter as a model provider backend.
• ›Adds a telemetry service for usage tracking.
• ›Updates the /bug slash command to point to the Qwen-Code issue tracker instead of the upstream repo.

Qwen Code v0.0.5-nightly.9 adds OpenRouter support, Qwen OAuth, systemPromptMappings config, and a telemetry service.

$ git clone --branch v0.0.5-nightly.9 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.5-nightly.9

• ›Adds systemPromptMappings configuration feature, enabling per-context system prompt overrides.
• ›Adds Qwen OAuth integration for authenticated sessions.
• ›Adds support for OpenRouter as a model provider backend.
• ›Adds a telemetry service for usage tracking.
• ›Updates the /bug command to point to the Qwen-Code repository.

Qwen Code v0.0.5-nightly.8 adds OAuth, OpenRouter support, systemPromptMappings config, and a telemetry service.

$ git clone --branch v0.0.5-nightly.8 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.5-nightly.8

• ›Adds systemPromptMappings configuration feature, enabling per-context system prompt customization.
• ›Adds Qwen OAuth integration for authentication flows.
• ›Adds support for OpenRouter as a model provider.
• ›Adds a telemetry service for usage tracking.
• ›Updates the /bug command to point to the Qwen-Code issue tracker.

Qwen Code v0.0.5 adds OpenRouter support, Qwen OAuth, systemPromptMappings config, and a telemetry service.

$ git clone --branch v0.0.5 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.5

• ›Adds systemPromptMappings configuration feature to map system prompts conditionally.
• ›Adds Qwen OAuth integration for authenticated access.
• ›Supports OpenRouter as a model provider backend.
• ›Adds a telemetry service for usage tracking.
• ›Updates /bug command to point to the Qwen-Code repository.

Qwen Code v0.0.5-nightly.7 adds OAuth login, OpenRouter support, systemPromptMappings config, and a telemetry service.

$ git clone --branch v0.0.5-nightly.7 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.5-nightly.7

• ›Adds systemPromptMappings configuration feature, enabling per-context system prompt overrides.
• ›Adds Qwen OAuth integration for authenticated access.
• ›Adds support for OpenRouter as a model provider.
• ›Adds a telemetry service for usage data collection.
• ›Updates the /bug command to point to the Qwen-Code repository.

Qwen Code v0.0.5-nightly.6 adds OpenRouter support, Qwen OAuth, systemPromptMappings config, and a telemetry service.

$ git clone --branch v0.0.5-nightly.6 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.5-nightly.6

• ›Adds systemPromptMappings configuration feature, enabling per-context system prompt customization.
• ›Adds Qwen OAuth integration for authentication flows.
• ›Adds support for OpenRouter as a model provider.
• ›Adds a telemetry service for usage tracking.
• ›Updates the /bug command to point to the Qwen-Code repository.

Qwen Code v0.0.5-nightly.5 adds OpenRouter support, systemPromptMappings config, and a telemetry service.

$ git clone --branch v0.0.5-nightly.5 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.5-nightly.5

• ›Adds systemPromptMappings configuration feature, allowing per-context system prompt overrides.
• ›Adds support for OpenRouter as a model provider backend.
• ›Adds a telemetry service for usage tracking.
• ›Updates the /bug command to point to the Qwen-Code repository.

Qwen Code alpha.14 adds systemPromptMappings config, OpenRouter support, and a telemetry service.

$ git clone --branch v0.0.1-alpha.14 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.0.1-alpha.14

• ›Adds systemPromptMappings configuration feature, allowing per-model or per-context system prompt overrides.
• ›Supports OpenRouter as a new model provider integration.
• ›Adds a telemetry service for usage tracking.
• ›Updates the /bug command to point to the Qwen-Code repository.

Zed v0.199.5 adds OpenAI vision support, settings profiles, edit-prediction toggle, and GitLab PR links.

$ git clone --branch v0.199.5 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.199.5

{
  "profiles": {
    "Streaming": {
      "agent_font_size": 20,
      "buffer_font_size": 20,
      "theme": "One Light",
      "ui_font_size": 20
    }
  }
}

• ›Adds workspace::ToggleEditPrediction action to toggle show_edit_predictions for all files via Edit Predictions menu -> All Files.
• ›Introduces experimental profiles key in settings to configure named settings profiles; activate one with settings profile selector: toggle.
• ›Adds vision support for OpenAI models in the agent.
• ›Adds option to create a branch from the default branch in the git branch picker.
• ›Adds 'View Pull Request' action when pushing to GitLab remotes.

Zed v0.198.4 makes collaboration server connection opt-in instead of automatic on sign-in.

$ git clone --branch v0.198.4 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.198.4

• ›Collaboration server connection is now manual opt-in — Zed no longer automatically connects to the collaboration server on sign-in.

Zed v0.198.3 adds GPT-5, GPT-5 mini, and GPT-5 nano as available AI models via OpenAI.

$ git clone --branch v0.198.3 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.198.3

• ›Adds GPT-5, GPT-5 mini, and GPT-5 nano as selectable AI models, configured via the OPENAI_API_KEY environment variable.

Zed v0.198.2 adds Git stash/pop in the panel, Vim :norm support, outline collapse, and a new expand_outlines_with_depth setting.

$ git clone --branch v0.198.2 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.198.2

{
  "expand_outlines_with_depth": 2
}

• ›Adds expand_outlines_with_depth setting to control how deep the outline tree is expanded by default when a file is opened.
• ›Adds editor: convert to sentence case command.
• ›Adds collapse/expand functionality to outline view entries.
• ›Adds Git stash and pop-stash actions accessible via a menu entry in the Git panel.
• ›Adds Vim :norm command support, accepting both Vim-style (<C-w>) and Zed-style (<ctrl-w>) modifier key syntax; multi-line execution uses multi-cursor (combinational) rather than sequential.

Zed v0.197.6 adds support for Claude Opus 4.1 as an AI model option.

$ git clone --branch v0.197.6 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.197.6

• ›Adds Claude Opus 4.1 as a supported model.

gemini-cli v0.1.21 adds VS Code & Zed IDE integration, MCP OAuth, loop detection, custom themes, and a new /ide command suite.

$ git clone --branch v0.1.21 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.21

$ echo '{"hideBanner": true}' >> ~/.gemini/settings.json

$ gemini --proxy http://proxy.corp.example.com:8080

$ gemini
# then inside the session:
/ide status
/ide install

• ›Introduces a VS Code companion extension (VSIX bundled with release) for IDE-aware sessions.
• ›Adds Zed editor integration.
• ›New /ide status and /ide install commands to manage IDE integration from the CLI.
• ›Streams notifications when the active file changes in IDE mode, threading the open file into model context.
• ›Adds MCP OAuth infrastructure (Part 1) for authenticated MCP server connections.

gemini-cli v0.1.19-nightly adds MCP tool filtering, extension listing, SVG support, Cloud Shell auth reuse, and startup warnings.

$ git clone --branch v0.1.19-nightly.250814.514e883a https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.19-nightly.250814.514e883a

$ gemini --allowed-mcp-server-names my-server,another-server

# In settings.json or .gemini/config.json
{
  "mcpServers": {
    "my-server": {
      "command": "npx my-mcp-server",
      "excludeTools": ["dangerous_tool", "verbose_tool"]
    }
  }
}

$ gemini --list-extensions

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are active at startup.
• ›Adds excludeTools and includeTools options to mcpServers config to filter individual MCP tools per server.
• ›Adds a command-line option to enable and list extensions.
• ›Adds .svg file support for inline content handling.
• ›Enables Gemini CLI to reuse the user's existing auth in Google Cloud Shell.

gemini-cli v0.1.20 adds --approval-mode flag and a new VS Code IDE integration.

$ git clone --branch v0.1.20 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.20

$ gemini --approval-mode auto

• ›Adds --approval-mode parameter to control how the CLI handles tool-call approvals without interactive prompts.
• ›Launches VS Code IDE integration, enabling gemini-cli to operate as a backend for the VS Code extension.

gemini-cli v0.1.19-nightly adds MCP server filtering flags, SVG support, extension listing, Cloud Shell auth reuse, and more.

$ git clone --branch v0.1.19-nightly.250813.9d023be1 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.19-nightly.250813.9d023be1

$ gemini --allowed-mcp-server-names shodan,nuclei

{
  "mcpServers": {
    "my-server": {
      "command": "npx my-mcp-server",
      "excludeTools": ["delete_resource", "shutdown"]
    }
  }
}

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are activated at startup.
• ›Adds excludeTools and includeTools options in mcpServers config to filter individual tools exposed by an MCP server.
• ›Adds a command-line option to enable and list extensions.
• ›Supports .svg files as input via the @file mechanism.
• ›Enables Gemini CLI to reuse the user's existing auth when running inside Google Cloud Shell.

gemini-cli v0.1.19 adds MCP management commands, fzf file completion, subagent architecture, /settings panel, and IDE diff support.

$ git clone --branch v0.1.19 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.19

$ /settings

• ›Adds gemini mcp add, gemini mcp remove, and gemini mcp list subcommands for managing MCP server configurations from the CLI.
• ›Adds MCP Roots support for scoping MCP server access.
• ›Detects tools with bad or cyclic schemas and automatically omits them with a warning rather than failing silently.
• ›Introduces Foundational Subagent Architecture, enabling agent-spawning workflows within gemini-cli.
• ›Adds @-command file path completion powered by fzf integration for faster, fuzzy file targeting.

gemini-cli v0.1.18-nightly adds MCP server filtering flags, SVG support, extension listing, Cloud Shell auth reuse, and per-server tool inclusion/exclusion.

$ git clone --branch v0.1.18-nightly.250812.26fe587b https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.18-nightly.250812.26fe587b

$ gemini --allowed-mcp-server-names filesystem,github

# In settings.json
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/tmp"],
      "includeTools": ["read_file", "list_directory"],
      "excludeTools": ["write_file"]
    }
  }
}

$ gemini --list-extensions

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are activated at startup.
• ›Adds excludeTools and includeTools fields to mcpServers config for per-server tool filtering.
• ›Adds a command-line option to enable and list extensions.
• ›Adds .svg file support for inline content handling.
• ›Enables Gemini CLI to reuse the user's existing auth in Google Cloud Shell.

gemini-cli v0.1.18-nightly adds MCP server filtering flags, SVG support, extension listing, Cloud Shell auth reuse, and per-server tool inclusion/exclusion.

$ git clone --branch v0.1.18-nightly.250811.2865a527 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.18-nightly.250811.2865a527

$ gemini --allowed-mcp-server-names my-corp-server,approved-scanner

# In settings.json mcpServers block:
{
  "mcpServers": {
    "my-corp-server": {
      "command": "npx my-corp-mcp",
      "includeTools": ["read_file", "list_dir"],
      "excludeTools": ["exec_shell"]
    }
  }
}

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are activated at startup.
• ›Adds excludeTools and includeTools options to mcpServers config for per-server tool filtering.
• ›Adds a command-line option to enable and list extensions.
• ›Adds .svg file support for @file context inclusion.
• ›Enables reuse of the user's existing auth token when running inside Google Cloud Shell.

gemini-cli v0.1.18-nightly adds MCP server filtering flags, SVG support, extension listing, Cloud Shell auth reuse, and per-server tool inclusion/exclusion.

$ git clone --branch v0.1.18-nightly.250810.c632ec8b https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.18-nightly.250810.c632ec8b

$ gemini --allowed-mcp-server-names my-server,another-server

# in settings.json
{
  "mcpServers": {
    "my-server": {
      "command": "npx",
      "args": ["my-mcp-server"],
      "includeTools": ["read_file", "list_dir"],
      "excludeTools": ["delete_file"]
    }
  }
}

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are loaded at startup.
• ›Adds excludeTools and includeTools fields in mcpServers config to control which tools are exposed per MCP server.
• ›Adds a command-line option to enable and list extensions.
• ›Supports .svg files as input via the @file syntax.
• ›Enables Gemini CLI to reuse the user's existing auth in Google Cloud Shell.

gemini-cli v0.1.17-nightly adds MCP server filtering flags, SVG support, extension listing, Cloud Shell auth reuse, and per-server tool inclusion/exclusion.

$ git clone --branch v0.1.17-nightly.250809.f35921a7 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.17-nightly.250809.f35921a7

$ gemini --allowed-mcp-server-names my-security-server

{
  "mcpServers": {
    "my-security-server": {
      "command": "npx",
      "args": ["my-mcp-server"],
      "includeTools": ["run_scan", "get_findings"],
      "excludeTools": ["delete_policy"]
    }
  }
}

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are activated at startup.
• ›Adds excludeTools and includeTools options in mcpServers config to control which tools each MCP server exposes.
• ›Adds a command-line option to enable and list extensions.
• ›Adds .svg file support for inline content handling.
• ›Enables Gemini CLI to reuse the user's existing auth in Google Cloud Shell.

gemini-cli v0.1.17-nightly adds MCP tool filtering, SVG support, Cloud Shell auth reuse, extension listing, and more.

$ git clone --branch v0.1.17-nightly.250808.60362e03 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.17-nightly.250808.60362e03

$ gemini --allowed-mcp-server-names=my-server,another-server

# In settings.json or .gemini/settings.json
{
  "mcpServers": {
    "my-server": {
      "command": "npx my-mcp-server",
      "includeTools": ["read_file", "list_dir"],
      "excludeTools": ["delete_file"]
    }
  }
}

$ gemini --list-extensions

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are active at startup.
• ›Adds excludeTools and includeTools options in mcpServers config to filter tools per MCP server.
• ›Adds a command-line option to enable and list extensions.
• ›Supports .svg files as input (via @file or context).
• ›Enables Gemini CLI to reuse the user's existing auth in Google Cloud Shell.

gemini-cli v0.1.18 adds VS Code/Zed IDE integration, MCP OAuth, loop detection, proxy support, and custom themes.

$ git clone --branch v0.1.18 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.18

# In your settings.json
{ "hideBanner": true }

$ gemini --proxy http://proxy.corp.example.com:8080 "Summarize the findings in report.md"

$ /ide install
/ide status

• ›Introduces a VS Code companion extension enabling IDE integration, with /ide status and /ide install commands to manage it.
• ›Adds Zed editor integration alongside VS Code IDE mode.
• ›Adds MCP OAuth infrastructure (Part 1) for authenticating MCP servers via OAuth.
• ›Introduces a loop detection service that automatically breaks simple agentic loops, with LLM-based loop checking as a second layer.
• ›Adds support for allowed/excluded MCP server name filtering in settings.

gemini-cli v0.1.17-nightly adds MCP tool filtering, extension listing, SVG support, Cloud Shell auth reuse, and startup warnings.

$ git clone --branch v0.1.17-nightly.250806.805114ae https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.17-nightly.250806.805114ae

$ gemini --allowed-mcp-server-names my-server,audit-server

{
  "mcpServers": {
    "my-server": {
      "command": "npx",
      "args": ["my-mcp-server"],
      "excludeTools": ["delete_database", "drop_table"]
    }
  }
}

$ gemini --list-extensions

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are active at launch.
• ›Adds excludeTools and includeTools fields to mcpServers config for fine-grained MCP tool filtering.
• ›Adds a command-line option to enable and list extensions.
• ›Adds .svg file support for inline content handling.
• ›Enables Gemini CLI to reuse the user's existing auth in Google Cloud Shell.

gemini-cli v0.1.17 adds IDE integration, MCP OAuth, loop detection, proxy support, custom themes, and new /ide commands.

$ git clone --branch v0.1.17 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.17

# In ~/.gemini/settings.json
{
  "hideBanner": true
}

$ gemini --proxy http://proxy.corp.example.com:8080

$ /ide install
/ide status

• ›Adds hideBanner setting to disable the startup banner.
• ›Introduces VSCode companion extension for IDE integration with gemini-cli.
• ›Adds /ide status and /ide install commands to manage IDE integration from the CLI.
• ›Streams notifications to the CLI when the active file changes in the IDE, and includes the active file in model requests.
• ›Introduces Zed editor integration.

gemini-cli v0.1.16-nightly adds MCP tool filtering, SVG support, Cloud Shell auth reuse, and a new extensions CLI flag.

$ git clone --branch v0.1.16-nightly.250805.99ba2f64 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.16-nightly.250805.99ba2f64

$ gemini --allowed-mcp-server-names my-server,audit-server

{
  "mcpServers": {
    "my-server": {
      "command": "npx",
      "args": ["my-mcp-server"],
      "includeTools": ["read_file", "list_dir"],
      "excludeTools": ["exec_shell"]
    }
  }
}

$ gemini --extensions

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are active at runtime.
• ›Adds excludeTools and includeTools fields to mcpServers config for per-server tool filtering.
• ›Adds a command-line option (--extensions) to enable and list extensions.
• ›Adds .svg file support for inline content handling.
• ›Enables Gemini CLI to reuse the user's existing auth when running inside Google Cloud Shell.

gemini-cli v0.1.16-nightly adds MCP tool filtering, SVG support, Cloud Shell auth reuse, and a new extensions CLI flag.

$ git clone --branch v0.1.16-nightly.250804.a8984a9b https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.16-nightly.250804.a8984a9b

$ gemini --allowed-mcp-server-names shodan,burp

# in settings.json
{
  "mcpServers": {
    "my-server": {
      "command": "my-mcp-server",
      "includeTools": ["read_file", "list_dir"]
    }
  }
}

$ gemini --extensions

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are active at launch.
• ›Adds excludeTools and includeTools per-server config options in mcpServers to whitelist or blacklist individual MCP tools.
• ›Adds a command-line option (--extensions) to enable and list extensions.
• ›Adds .svg file support for inline content input.
• ›Enables Gemini CLI to reuse the user's existing auth in Google Cloud Shell.

gemini-cli v0.1.15-nightly adds MCP server filtering flags, SVG support, extension listing, Cloud Shell auth reuse, and per-server tool inclusion/exclusion.

$ git clone --branch v0.1.15-nightly.250803.820169ba https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.15-nightly.250803.820169ba

$ gemini --allowed-mcp-server-names my-server,audit-server

# In settings.json
{
  "mcpServers": {
    "my-server": {
      "command": "npx my-mcp-server",
      "excludeTools": ["dangerous_tool", "write_file"]
    }
  }
}

$ gemini --list-extensions

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers the CLI connects to at startup.
• ›Adds excludeTools and includeTools config options per mcpServers entry to fine-tune which MCP tools are exposed.
• ›Adds a command-line option to enable and list extensions.
• ›Adds .svg file support for inline content handling.
• ›Enables reuse of the user's existing auth in Google Cloud Shell, avoiding re-authentication.

gemini-cli v0.1.16 adds IDE integration with VS Code & Zed, MCP OAuth, loop detection, custom themes, and proxy support.

$ git clone --branch v0.1.16 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.16

# In your settings.json:
{ "hideBanner": true }

$ gemini --proxy http://proxy.corp.example.com:8080

$ /ide status
/ide install

• ›Adds hideBanner setting to disable the startup banner in settings.json.
• ›Introduces VS Code companion extension for IDE integration with gemini-cli.
• ›Adds Zed editor integration.
• ›Adds /ide status and /ide install commands to manage IDE integration from the CLI.
• ›Shows the active file open in the IDE in the context section above the input box during IDE sessions.

gemini-cli v0.1.15-nightly adds MCP server filtering flags, SVG support, extension listing, Cloud Shell auth reuse, and more.

$ git clone --branch v0.1.15-nightly.250802.15a1f1af https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.15-nightly.250802.15a1f1af

$ gemini --allowed-mcp-server-names my-server,audit-server

# In settings.json or .gemini/settings.json
{
  "mcpServers": {
    "my-server": {
      "command": "npx",
      "args": ["my-mcp-server"],
      "excludeTools": ["run_shell", "delete_file"]
    }
  }
}

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are active at launch.
• ›Adds excludeTools and includeTools per-server options in mcpServers config to filter individual MCP tools.
• ›Adds .svg file support for inline content input.
• ›Adds a command-line option to enable and list extensions.
• ›Enables Gemini CLI to reuse the user's existing auth when running inside Google Cloud Shell.

gemini-cli v0.1.15-nightly adds MCP server filtering flags, SVG support, extension listing, Cloud Shell auth reuse, and per-server tool include/exclude controls.

$ git clone --branch v0.1.15-nightly.250801.6f7beb41 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.1.15-nightly.250801.6f7beb41

$ gemini --allowed-mcp-server-names my-safe-server,audit-server

# In settings.json or .gemini/config.json
{
  "mcpServers": {
    "my-server": {
      "command": "npx my-mcp-server",
      "includeTools": ["read_file", "list_dir"]
    }
  }
}

$ gemini --list-extensions

• ›Adds --allowed-mcp-server-names flag to restrict which MCP servers are loaded at startup.
• ›Adds excludeTools and includeTools per-server config keys in mcpServers to control which tools each MCP server exposes.
• ›Adds a command-line option to enable and list extensions.
• ›Adds .svg file support for inline content.
• ›Enables Gemini CLI to reuse the user's existing auth in Google Cloud Shell.

LocalAI v3.3.2 lets backends install from a local path and run without a metadata file.

$ git clone --branch v3.3.2 https://github.com/mudler/LocalAI.git
# already have the repo? check out this version:
$ git checkout v3.3.2

• ›Supports installing backends from a local filesystem path, enabling offline or custom backend deployments.
• ›Allows backends to operate without a metadata file, removing a previously required dependency for backend registration.

LocalAI v3.3.1 adds Flux Kontext image editing via ref_images API field and LoRA loading for stable-diffusion-ggml.

$ git clone --branch v3.3.1 https://github.com/mudler/LocalAI.git
# already have the repo? check out this version:
$ git checkout v3.3.1

$ local-ai run flux.1-kontext-dev

• ›Adds ref_images field to the POST /v1/images/generations API, enabling Flux Kontext-powered image editing by passing one or more reference image URLs alongside a prompt.
• ›Adds support for loading LoRAs in the stable-diffusion-ggml backend.
• ›Adds flux.1-kontext-dev model, launchable with local-ai run flux.1-kontext-dev, for in-context image editing.
• ›Adds new models to the model gallery: flux.1-krea-dev-ggml, flux.1-dev-ggml-q8_0, flux.1-dev-ggml-abliterated-v2-q8_0, qwen_qwen3-30b-a3b-instruct-2507, qwen_qwen3-30b-a3b-thinking-2507, and arcee-ai_afm-4.5b.

GPT-OSS template 'Continue' and 'Impersonate' actions now work correctly, plus a llama.cpp update.

$ git clone --branch v3.9.1 https://github.com/oobabooga/textgen.git
# already have the repo? check out this version:
$ git checkout v3.9.1

• ›GPT-OSS template special actions 'Continue' and 'Impersonate' now function correctly.

oobabooga textgen v3.9 adds experimental GPT-OSS support with reasoning effort controls and standalone Jinja chat templates.

$ git clone --branch v3.9 https://github.com/oobabooga/textgen.git
# already have the repo? check out this version:
$ git checkout v3.9

• ›Adds a Reasoning effort UI element in the chat tab with low, medium, and high options for GPT-OSS models.
• ›Supports standalone .jinja chat templates, enabling GPT-OSS to be loaded through the Transformers backend.
• ›Makes web search functional with thinking models.
• ›Experimental support for GPT-OSS GGUF models (20B and 120B variants) via updated llama.cpp and Transformers 4.55.

Ollama v0.11.4 improves OpenAI-compatible API support: simultaneous content+tool calls, tool name propagation, and always-on reasoning.

$ git clone --branch v0.11.4 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.11.4

• ›Supports content and tool calls coexisting in the same OpenAI-compatible message, enabling richer agentic responses.
• ›Propagates the tool name when converting role=tool messages in the OpenAI-compatible layer.
• ›Always provides reasoning output in OpenAI-compatible responses.

Ollama v0.11 adds native support for OpenAI's gpt-oss 20B and 120B open-weight models with built-in web search and MXFP4 quantization.

$ git clone --branch v0.11.0 https://github.com/ollama/ollama.git
# already have the repo? check out this version:
$ git checkout v0.11.0

$ ollama run gpt-oss:20b

$ ollama run gpt-oss:120b

• ›Adds OpenAI gpt-oss 20B and 120B open-weight models, runnable locally via ollama run gpt-oss:20b and ollama run gpt-oss:120b.
• ›Supports native MXFP4 quantization format for gpt-oss MoE weights (4.25 bits/param), enabling the 20B to run on 16 GB RAM and the 120B on a single 80 GB GPU.
• ›Enables built-in optional web search to augment gpt-oss models with real-time information.
• ›Supports configurable reasoning effort (low, medium, high) for gpt-oss models to balance quality and latency.
• ›Exposes full chain-of-thought reasoning output from gpt-oss models.