The daily firehose — everything the toolchain shipped today, already filtered.
// HOW THIS ISSUE IS MADE
We read every release from the 174 tools on our watchlist at the source — GitHub and GitLab release notes, vendor release pages and changelogs, project blogs and feeds, vendor press releases, and the source code behind the tag. Bug-fix-only releases and non-product newsroom noise are dropped; what's left is summarized down to the new capability, how to try it, and any screenshots or videos the release itself published. Every entry links to the sources it was built from.
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 v2.1.224 adds self-hosted runners, archive plugin installs, JWT/AWS credential masking, and cross-session messaging.
└──▷ GET THIS VERSION
$ git clone --branch v2.1.224 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:$ git checkout v2.1.224
└──▷ TRY IT
Stand up a self-hosted runner so your team's Claude Code web and mobile sessions execute on your own infrastructure instead of Anthropic-managed compute.
$ claude self-hosted-runner
Route Bedrock inference to a specific cross-region profile (e.g. us-west-2) regardless of the ambient AWS_REGION, useful when your deployment region differs from the inference profile region.
$ export ANTHROPIC_BEDROCK_REGION_PREFIX=us-west-2
claude
›Adds claude self-hosted-runner subcommand to turn your own machines or containers into execution targets for Claude Code web, mobile, and desktop sessions (Team and Enterprise plans).
›Adds archive plugin source to install plugins from a zip over HTTPS without git or npm, with optional SHA-256 pinning.
›Adds ANTHROPIC_BEDROCK_REGION_PREFIX environment variable to prefer a specific cross-region Bedrock inference profile over the AWS_REGION-derived one.
›Adds crossSessionInbound and dialogExpiry settings: cross-session messages to bypassed-permission sessions are held for approval; messages to other sessions auto-deliver.
›Adds sandbox credential-masking options extract, onExtractNoMatch, decode: "jwt" with maskClaims, and awsPairs/sigv4 for AWS SigV4 re-signing; requires network.tlsTerminate and honored only from user, managed, or --settings settings.
+7 moreshow less
›Adds cross-session SendMessage and ListAgents APIs: Claude Code sessions on macOS and Linux can now discover and message each other across machines.
›Removes the 200-subagent-per-session spawn cap, allowing long-running sessions to continue launching new agents (concurrency and depth limits still apply).
›Adds a cancel-and-confirm step when removing an unavailable paste changes a command's text.
›Improves fullscreen mode to retain the full pre-compaction history in scrollback across repeated compactions, not just the most recent interval.
›Improves Remote Control: attached web and mobile clients now see compaction progress and the post-compaction boundary, and /clear resets propagate to attached clients.
›Improves Remote Control: connection failures now show a persistent failure indicator with details and a reconnect shortcut instead of an 8-second toast.
›Changes feedback-survey transcript sharing to also upload (with consent) the last request's model settings — system prompt (including CLAUDE.md), tool definitions, and model parameters — with secrets redacted.
Autonomous coding agent as an SDK, IDE extension, or CLI assistant.
Cline desktop v0.0.10 adds OAuth for remote MCP servers, new keyboard shortcuts, and improved reasoning controls across AI SDK providers.
└──▷ GET THIS VERSION
$ git clone --branch desktop-v0.0.10 https://github.com/cline/cline.git
# already have the repo? check out this version:$ git checkout desktop-v0.0.10
›New keyboard shortcuts: Cmd/Ctrl+N for a new session and Cmd/Ctrl+, for settings; Esc now stops the current turn.
›Remote MCP servers can now authenticate with OAuth via Settings → MCP — authorize a server, view its auth status, and cancel or retry a pending authorization.
›Supports MCP servers that require a pre-registered OAuth client (client ID/secret) instead of dynamic registration; stored tokens are invalidated when a server's client configuration changes.
›MCP errors are now surfaced per-server rather than as a page-level error, and servers with invalid configuration display their error inline instead of silently disappearing.
›Failed turns now show an error in the transcript with the underlying cause and a pointer to Settings → Models when no model credentials are present.
+3 moreshow less
›Adds a 'Connect a model' notice on the welcome screen when no provider has credentials, with one-click access to onboarding or model settings; reacts live as credentials are added and recognizes Bedrock/Vertex and keyless local endpoints as already connected.
›Reasoning controls now resolve consistently across AI SDK providers including Ollama, so effort levels and thinking on/off are honored wherever the provider supports them.
›Vertex AI credential refreshes now use the configured fetch, enabling ADC authentication behind proxies and custom networking setups.
$ git clone --branch rust-v0.148.0-alpha.2 https://github.com/openai/codex.git
# already have the repo? check out this version:$ git checkout rust-v0.148.0-alpha.2
└──▷ TRY IT
Fork an existing session by ID and immediately send a follow-up prompt to continue work on a new branch of the conversation.
$ codex exec fork session-123 'continue refactoring the auth module'
Export the current TUI conversation to a Markdown file for documentation or sharing.
$ /export ~/projects/my-session.md
Point agent identity registration and verification at a staging environment instead of production.
›Adds fork subcommand to codex exec that forks a previous session by UUID or thread name into a new session, with optional --image/-i flag for attachments and an optional prompt argument.
›Adds /export TUI command to export the complete conversation history as structured Markdown, with clipboard and file destinations (including default filename prompt and support for an explicit path argument).
›Supports CODEX_AGENT_IDENTITY_AUTHAPI_BASE_URL and CODEX_AGENT_IDENTITY_JWKS_BASE_URL environment variables to override agent identity registration and verification endpoints.
›Adds session archiving to the resume picker, allowing sessions to be archived and restored from within the TUI.
›Recognizes mcp_tool hook handler configurations (with server, tool, JSON input, and optional timeout and status message fields) and returns them from configRequirements/read.
+1 moreshow less
›Adds executor-local config layer loading for the exec server, enabling per-executor configuration overrides.
Codex CLI adds migrate-rollouts subcommand, Guardian circuit-breaker for cyber models, and a durable user-message queue.
└──▷ GET THIS VERSION
$ git clone --branch rust-v0.148.0-alpha.1 https://github.com/openai/codex.git
# already have the repo? check out this version:$ git checkout rust-v0.148.0-alpha.1
└──▷ TRY IT
Dry-run to see which local sessions are eligible for migration to paginated thread history before committing.
$ codex migrate-rollouts --verbose--json
Migrate a specific thread to paginated history with a bandwidth cap to avoid saturating disk I/O.
›Adds migrate-rollouts subcommand with --apply, --thread <THREAD_ID>, --max-mib-per-second <MIB>, --json, and --verbose flags to inspect or migrate legacy local sessions to paginated thread history.
›Adds a Guardian circuit-breaker policy that interrupts model turns after the first Guardian denial for models whose catalog specialty is cyber, while retaining existing denial thresholds for all other models.
›Mounts a minimal /dev inside full-filesystem Bubblewrap sandboxes.
›Starts cached MCP servers lazily for subagents, and recovers MCP servers automatically after OAuth reauthentication.
›Fetches remote installed plugins across all scopes and supports plugin roots in the host skill loader.
+1 moreshow less
›Sends model routing hints to the Codex backend to improve model selection.
OpenCode v1.18.15 adds session JSON export from the UI and broader desktop locale coverage.
└──▷ GET THIS VERSION
$ git clone --branch v1.18.15 https://github.com/anomalyco/opencode.git
# already have the repo? check out this version:$ git checkout v1.18.15
›Adds Export... option to the 3-dot session dropdown menu, an Export session button in the Context tab, and a session.export command palette action (/export) to export full session transcripts as JSON from the desktop UI.
›Adds much broader locale coverage across the desktop app.
Pi v0.84.1 adds pi auth check for credential preflight, Qwen Token Plan Individual provider, and extension terminate support for tool call batches.
└──▷ GET THIS VERSION
$ git clone --branch v0.84.1 https://github.com/earendil-works/pi.git
# already have the repo? check out this version:$ git checkout v0.84.1
└──▷ TRY IT
Run a credential preflight check before kicking off a long agent session to confirm your provider keys are valid.
$ pi auth check
›Adds pi auth check subcommand to verify provider or model credentials before a run, with optional output of the resolved credential.
›Adds Qwen Token Plan Individual as a built-in provider, using the shared international QWEN_TOKEN_PLAN_API_KEY environment variable and its documented subscription model catalog.
›Adds terminate support to blocked extension tool_call event handlers, allowing all-terminating batches to skip the automatic follow-up model call.
›Adds half-page transcript scrolling actions (unbound keybindings) for the TUI fullscreen viewport.
›Adds word/whitespace double-click selection, granularity-aware drag selection, and triple-click paragraph selection in fullscreen mode.
LangGraph checkpoint 4.2.0 adds opt-in omit_expired flag to skip expired checkpoint rows on read.
└──▷ GET THIS VERSION
$ git clone --branch checkpoint==4.2.0 https://github.com/langchain-ai/langgraph.git
# already have the repo? check out this version:$ git checkout checkpoint==4.2.0
›Adds opt-in omit_expired parameter to checkpoint and checkpoint-postgres readers to skip expired rows when reading checkpoint history, reducing noise and improving read performance in long-running workflows.
LangGraph checkpoint 4.2.0 adds opt-in omit_expired flag to skip expired rows on checkpoint reads.
└──▷ GET THIS VERSION
$ git clone --branch checkpoint==4.2.0 https://github.com/langchain-ai/langgraph.git
# already have the repo? check out this version:$ git checkout checkpoint==4.2.0
›Adds omit_expired opt-in parameter to checkpoint and checkpoint-postgres read operations to skip expired rows, enabling cleaner state retrieval without stale data.
PydanticAI v2.26.0 adds run cancellation, hidden/revealed tools, DeepSeek V4 Flash support, and a public AgentRunEvents handle.
└──▷ GET THIS VERSION
$ git clone --branch v2.26.0 https://github.com/pydantic/pydantic-ai.git
# already have the repo? check out this version:$ git checkout v2.26.0
└──▷ USE IT
Stream agent events and cancel mid-run based on application logic using the new public AgentRunEvents handle.
python
async with agent.run_stream_events(user_prompt) as events:
async for event in events:
if should_stop(event):
events.cancel()
break
Use DeepSeek V4 Flash in an agent via the OpenAI-compatible responses model and DeepSeek provider.
python
from pydantic_ai.models.openai import OpenAIResponsesModel
from pydantic_ai.providers.deepseek import DeepSeekProvider
model = OpenAIResponsesModel('deepseek-chat', provider=DeepSeekProvider())
agent = Agent(model=model)
›Adds AgentRun.cancel() and RunContext.cancel() for first-party run cancellation, raising RunCancelled to stop in-flight agent runs programmatically.
›Adds Model.resolve_prompt_cache_retention() to resolve the effective prompt-cache retention setting from model settings.
›Promotes the run_stream_events() iterator to a public AgentRunEvents handle exposing cancel() and run-state access.
›Supports hiding function tools until revealed — via tool search, load_capability, or ToolReturn.tools — using each provider's native deferral/addition channel.
›Covers DeepSeek V4 Flash via OpenAIResponsesModel and DeepSeekProvider.
$ git clone --branch b10305 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:$ git checkout b10305
›Adds SYCL backend support for four DSv4 operations: LIGHTNING_INDEXER, DSV4_HC_COMB, DSV4_HC_POST, and DSV4_HC_PRE, expanding hardware-accelerated inference coverage on Intel SYCL devices.
An MCP that lets AI tools securely connect to your infrastructure, write IaaS code, debug issues, and assist during incidents - without risking production stability. Built for security teams to approve and infrastructure teams to experience like magic.
$ git clone --branch runner-v0.17.2 https://github.com/AndrewDryga/emisar.git
# already have the repo? check out this version:$ git checkout runner-v0.17.2
›Adds BunnyCDN operations and explicit runbook targets.
›Allows production MIG rollout placement in available zones.
Phoenix v19.19.0 collapses older messages in LLM span views to reduce noise.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v19.19.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-v19.19.0
›In the trace UI, all messages except the last are now collapsed by default in LLM spans, reducing visual clutter when inspecting long conversations.
THE AI TOOLCHAIN