The AI Toolchain — issue -133, April 8, 2026

AutoGPT Platform v0.6.54 adds dry-run agent loops, extended thinking, SQL analytics, /orchestrate fleet supervisor, and 12 new GLM models via OpenRouter.

$ git clone --branch autogpt-platform-beta-v0.6.54 https://github.com/Significant-Gravitas/AutoGPT.git
# already have the repo? check out this version:
$ git checkout autogpt-platform-beta-v0.6.54

• ›Adds dry_run session-level flag to autopilot sessions, enabling a create → dry-run → fix loop for agent generation without live execution.
• ›Adds include_graph option to the find_library_agent API call, letting callers retrieve the full agent graph alongside library results.
• ›Adds extended thinking execution mode to OrchestratorBlock for deeper multi-step reasoning.
• ›Adds a SQL query block with multi-database support for CoPilot analytics.
• ›Adds /orchestrate skill — a Claude Code agent fleet supervisor with spare worktree lifecycle management.

Framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI empowers agents to work together seamlessly, tackling complex tasks.

CrewAI 1.14.1 adds async checkpoint TUI browser and proper resource management for streaming outputs.

$ git clone --branch 1.14.1 https://github.com/crewAIInc/crewAI.git
# already have the repo? check out this version:
$ git checkout 1.14.1

• ›Adds async checkpoint TUI browser for interactive inspection of crew checkpoints.
• ›Adds aclose()/close() methods and async context manager support to streaming outputs for proper resource cleanup.

The agent that grows with you

Hermes v2026.4.8 adds MCP OAuth 2.1, live model switching, background task notifications, Google AI Studio, and OSV malware scanning.

$ git clone --branch v2026.4.8 https://github.com/NousResearch/hermes-agent.git
# already have the repo? check out this version:
$ git checkout v2026.4.8

$ /model gemini-2.5-pro

$ hermes logs --tail

• ›Adds notify_on_complete for background tasks — agent is automatically notified when long-running processes (builds, tests, deployments) finish, eliminating polling.
• ›New /model command enables live model and provider switching mid-session across CLI, Telegram, Discord, Slack, and other gateway platforms.
• ›Adds native Google AI Studio (Gemini) provider with automatic context length detection via models.dev registry.
• ›Adds MCP OAuth 2.1 PKCE support for standards-compliant MCP server authentication.
• ›Adds automatic OSV malware scanning of MCP extension packages on install.

Build resilient agents.

LangGraph CLI gains remote build support for langgraph deploy

$ git clone --branch cli==0.4.20 https://github.com/langchain-ai/langgraph.git
# already have the repo? check out this version:
$ git checkout cli==0.4.20

• ›Adds remote build support for langgraph deploy, enabling builds to run on remote infrastructure instead of locally.

Your own personal AI assistant. Any OS.

OpenClaw v2026.4.7 adds a first-class infer hub, pluggable compaction, webhook ingress, memory-wiki, Arcee AI, and Gemma 4 support.

$ git clone --branch v2026.4.7 https://github.com/openclaw/openclaw.git
# already have the repo? check out this version:
$ git checkout v2026.4.7

agents:
  defaults:
    compaction:
      provider: my-custom-compaction-plugin

agents:
  defaults:
    systemPromptOverride: "You are a security analyst. Focus only on CVE triage."

• ›Auto-fallback across image, music, and video providers with intent preservation and capability remapping for size/aspect/resolution/duration hints.
• ›New bundled webhook ingress plugin enables external automation to create and drive bound TaskFlows through per-route shared-secret endpoints.
• ›Pluggable compaction provider registry lets plugins replace the built-in summarization pipeline, configured via agents.defaults.compaction.provider.
• ›New agents.defaults.systemPromptOverride for controlled prompt experiments plus heartbeat prompt-section controls.
• ›Gemma 4 model support added for the Google provider, including reasoning support and explicit thinking-off semantics in compatibility wrappers.

PydanticAI v1.78.0 adds return_schema, function_signature, and SetToolMetadata to ToolDefinition, plus OTel cached token span attributes.

$ git clone --branch v1.78.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:
$ git checkout v1.78.0

from pydantic_ai.tools import ToolDefinition

def my_tool(x: int) -> str:
    return str(x)

td = ToolDefinition(
    name='my_tool',
    description='Converts int to str',
    parameters_json_schema={},
    return_schema=...,           # new field
    function_signature=...,      # new field
)
print(td.return_schema)
print(td.function_signature)

• ›Adds return_schema and function_signature fields to ToolDefinition, exposing richer tool metadata for inspection and downstream use.
• ›Adds SetToolMetadata capability, enabling dynamic mutation of tool metadata at runtime.
• ›Adds cached token span attributes to OTel traces per the OpenTelemetry specification, improving observability of token usage.

Semantic Kernel Python 1.41.2 adds FunctionChoiceBehavior filter support in OpenAI responses agent tools.

$ git clone --branch python-1.41.2 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:
$ git checkout python-1.41.2

• ›Applies FunctionChoiceBehavior filters when building tool lists for OpenAI responses agents, ensuring function inclusion/exclusion rules are respected at invocation time.

Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.

Claude Code v2.1.97 adds focus view toggle, refreshable status lines, git worktree context, live subagent indicators, and Cedar policy syntax highlighting.

$ git clone --branch v2.1.97 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:
$ git checkout v2.1.97

# In settings.json
{
  "statusLine": {
    "command": "date '+%H:%M:%S'",
    "refreshInterval": 5
  }
}

# Example status-line script that reads workspace.git_worktree from stdin
import sys, json
data = json.load(sys.stdin)
worktree = data.get("workspace", {}).get("git_worktree", "")
print(f"worktree: {worktree}" if worktree else "main worktree")

$ # While Claude Code is running in NO_FLICKER mode, press:
Ctrl+O

• ›Adds focus view toggle (Ctrl+O) in NO_FLICKER mode showing prompt, one-line tool summary with edit diffstats, and final response
• ›Adds refreshInterval status line setting to re-run the status line command every N seconds
• ›Adds workspace.git_worktree field to status line JSON input, populated when inside a linked git worktree
• ›Adds ● N running indicator in /agents next to agent types with live subagent instances
• ›Adds syntax highlighting for Cedar policy files (.cedar, .cedarpolicy)

Glamourous agentic coding for all

Crush v0.56.0 adds crush_info and crush_logs self-diagnostic tools so the AI can inspect its own config and logs.

$ git clone --branch v0.56.0 https://github.com/charmbracelet/crush.git
# already have the repo? check out this version:
$ git checkout v0.56.0

{
  "options": {
    "disabled_tools": ["crush_info", "crush_logs"]
  }
}

• ›Adds crush_info tool: exposes the final merged configuration to the model so it can self-diagnose provider, LSP, and MCP setup.
• ›Adds crush_logs tool: surfaces Crush's own application logs to the model for in-session debugging.
• ›Supports disabling either self-diagnostic tool via the disabled_tools config option to prevent the model from ever calling them.
• ›Automatically opens the Hyper auth dialog on unauthorized errors, reducing manual intervention when credentials expire.

an open source, extensible AI agent that goes beyond code suggestions - install, execute, edit, and test with any LLM

Goose v1.30.0 adds goose serve, goose doctor, egress logging, new Gemini/Copilot/Zhipu providers, and major TUI improvements.

$ git clone --branch v1.30.0 https://github.com/block/goose.git
# already have the repo? check out this version:
$ git checkout v1.30.0

$ goose serve

$ goose doctor

$ GOOSE_SHOW_FULL_OUTPUT=true goose run --text 'List all open ports and their processes'

• ›New goose serve subcommand runs Goose as a background service.
• ›New goose doctor command diagnoses common configuration and environment issues.
• ›New egress logging inspector logs and inspects all outbound network calls for security visibility.
• ›Adds Gemini OAuth provider for first-class OAuth-based authentication with Google Gemini.
• ›Adds Copilot ACP provider for GitHub Copilot via ACP.

The open source coding agent.

OpenCode v1.4.0 adds OTLP observability, HTTP proxy support, PDF drag-and-drop, and a new --dangerously-skip-permissions run flag.

$ git clone --branch v1.4.0 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.4.0

$ opencode run --dangerously-skip-permissions

• ›Adds OTLP observability export support for tracing and monitoring OpenCode sessions.
• ›Adds full HTTP proxy support for routing OpenCode traffic through a proxy.
• ›Adds opencode run --dangerously-skip-permissions flag to auto-approve non-denied permission prompts.
• ›Adds a configurable keybinding for the 'Switch model variant' command in the TUI.
• ›Adds PDF drag-and-drop support for attachments in the TUI.

AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI

Pi v0.66.0 adds an Earendil startup announcement with inline image rendering and an Anthropic subscription billing warning.

$ git clone --branch v0.66.0 https://github.com/earendil-works/pi.git
# already have the repo? check out this version:
$ git checkout v0.66.0

• ›Displays an Earendil startup announcement with bundled inline image rendering and a linked blog post during interactive mode on April 8 and 9, 2026.
• ›Shows an interactive warning when Anthropic subscription auth is active, clarifying that third-party Anthropic usage draws from extra usage and is billed per token.

Qwen Code v0.14.2 adds /plan command, thinking block retention, adaptive token escalation, and a test-engineer agent.

$ git clone --branch v0.14.2 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:
$ git checkout v0.14.2

$ /plan

• ›Adds /plan command to the CLI for entering plan mode.
• ›Adds thinking block cross-turn retention with idle cleanup, preserving reasoning context across conversation turns.
• ›Adds adaptive output token escalation: defaults to 8K tokens and automatically retries at 64K when needed.
• ›Adds bugfix workflow, test-engineer agent, and debugging skills to the agent toolkit.
• ›Adds qwen3.6-plus model to the ModelStudio Coding Plan.

Zed v0.231.1 adds top-down agent streaming, Git Graph, native devcontainer support with Zed extensions, and configurable diff width switching.

$ git clone --branch v0.231.1 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:
$ git checkout v0.231.1

"agent": {
  "thinking_display": "always_expanded"
}

"minimum_split_diff_width": 120

{
  "customizations": {
    "zed": {
      "extensions": ["rust-analyzer", "toml"]
    }
  }
}

• ›Adds agent.thinking_display setting to control thinking block display in agent threads, accepting values 'automatic', 'always_expanded', or 'always_collapsed'.
• ›Adds minimum_split_diff_width setting to configure the column threshold at which the git diff view automatically switches from split mode to unified mode.
• ›Adds wrap_around option to pane::ActivatePreviousItem and pane::ActivateNextItem actions to optionally disable tab-cycling wrap-around behavior.
• ›Adds customizations.zed.extensions field in devcontainer.json to declare Zed extensions in devcontainers, backed by a new native devcontainer implementation replacing the previous Node-based CLI.
• ›Adds Git Graph view, accessible via the button at the bottom of the git panel or the git graph: Open action.

An open-source AI agent that brings the power of Gemini directly into your terminal.

gemini-cli v0.37.0 adds sandbox lockdown, persistent browser sessions, cross-platform notifications, new skills, and more.

$ git clone --branch v0.37.0 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:
$ git checkout v0.37.0

• ›Adds maxActionsPerTask setting for the browser agent to cap the number of actions per task.
• ›Adds a CI skill for automated failure replication.
• ›Implements forbiddenPaths for OS-specific sandbox managers to block access to sensitive filesystem paths.
• ›Supports inline agentCardJson string literal for configuring remote agents without a separate file.
• ›Adds experiment-gated support for Gemini Flash 3.1 Lite model.

llama.cpp b8705 adds support for the Step3-VL-10B vision-language model.

$ git clone --branch b8705 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b8705

• ›Adds support for the Step3-VL-10B multimodal vision-language model, including fused QKV, tensor mapping via tensor_mapping.py, and understand_projector_stride read from the global config.

$ git clone --branch b8703 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b8703

Chroma 1.5.7 adds getCollectionById across all SDKs, streaming S3 uploads, log sharding, and stdout-only tracing.

$ git clone --branch 1.5.7 https://github.com/chroma-core/chroma.git
# already have the repo? check out this version:
$ git checkout 1.5.7

• ›Adds getCollectionById API across all client SDKs and the server, enabling collection lookup by ID directly.
• ›Adds put_stream to chroma-storage for streaming S3 uploads.
• ›Adds shard_index, num_shards, and log_upper_bound_offset parameters passed through to query and orchestrator, enabling log partitioning across active and non-active shards.
• ›Adds SegmentWriter and Flusher abstractions over shards for internal write path.
• ›Enables stdout-only tracing mode for simplified observability output.

Langfuse v3.166.0 adds a fields parameter to the GetTraceById API endpoint for selective field retrieval.

$ git clone --branch v3.166.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v3.166.0

• ›Adds fields parameter to the GetTraceById API endpoint, allowing callers to request only specific fields and reduce response payload size.

$ git clone --branch v3.165.0 https://github.com/langfuse/langfuse.git
# already have the repo? check out this version:
$ git checkout v3.165.0