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.
›Adds OpenAILikeEmbedder class for providers with OpenAI-compatible embedding endpoints (e.g. LiteLLM proxy).
›Adds a search_past_sessions + read_past_session two-step pattern so agents and teams can browse previous sessions, with num_past_sessions_to_search and num_past_session_runs_in_search to control search scope and preview depth.
›Adds num_runs parameter to read_past_session so the model can fetch a subset of turns from long sessions instead of pulling full conversation history.
›Session previews in search_past_sessions now show per-run user/assistant pairs instead of a single message.
AutoGPT Platform adds file upload to copilot chat, E2B cloud sandbox, multi-step browser automation, and MCP server discovery in Otto.
└──▷ GET THIS VERSION
$ git clone --branch autogpt-platform-beta-v0.6.50 https://github.com/Significant-Gravitas/AutoGPT.git
# already have the repo? check out this version:$ git checkout autogpt-platform-beta-v0.6.50
›Enables OpenRouter broadcast for the SDK /messages endpoint, expanding model routing options.
›Adds file upload support to copilot chat, allowing users to attach files within the chat interface.
›Introduces E2B cloud sandbox with unified file tools, persistent execution, and output truncation.
›Adds agent-browser multi-step browser automation tools for automated web interactions.
›Adds MCP server discovery and execution in Otto, enabling dynamic tool server integration.
+4 moreshow less
›Replaces the legacy copilot with a baseline non-SDK mode that includes tool calling support.
›Adds Langfuse SDK tracing for the Claude Agent SDK execution path.
›Adds text-to-speech and share output actions to the platform UI.
›Renders context compaction as tool-call UI events for improved observability.
langchain-openrouter 0.1.0 adds streaming token usage, cost metadata, default headers, and a standard model property.
└──▷ GET THIS VERSION
$ git clone --branch langchain-openrouter==0.1.0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:$ git checkout langchain-openrouter==0.1.0
└──▷ USE IT
Inspect per-call cost and token usage after an OpenRouter completion to track spend in production.
python
from langchain_openrouter import ChatOpenRouter
llm = ChatOpenRouter(model="openai/gpt-4o")
result = llm.invoke("Summarize the OWASP Top 10")
print(result.response_metadata["cost"])
print(result.response_metadata["cost_details"])
Stream a response and receive token usage counts incrementally for budget-aware pipelines.
python
from langchain_openrouter import ChatOpenRouter
llm = ChatOpenRouter(model="openai/gpt-4o", stream_usage=True)
for chunk in llm.stream("List common lateral movement techniques"):
print(chunk)
›Adds cost and cost_details fields to response_metadata on OpenRouter chat model responses, exposing per-call cost information.
›Adds streaming token usage support to the OpenRouter integration, making token counts available during streamed completions.
›Adds a model standard property to the OpenRouter (and Fireworks/Groq) chat model classes, aligning with the LangChain standard model interface.
›Adds default headers support to the OpenRouter integration, allowing custom HTTP headers to be set on every request.
Letta 0.16.6 expands the Conversations API with agent-direct mode, adds GPT-5.3 models, and raises context and memory limits.
└──▷ GET THIS VERSION
$ git clone --branch 0.16.6 https://github.com/letta-ai/letta.git
# already have the repo? check out this version:$ git checkout 0.16.6
└──▷ TRY IT
Send a message directly to a known agent without managing a separate conversation ID, using the new agent-direct mode.
$ POST /v1/conversations/default/messages
{
"agent_id": "<agent_id>",
"messages": [{"role": "user", "content": "Summarize my tasks."}]
}
›Adds conversation_id="default" + agent_id parameter support across conversation endpoints (send, list, cancel, compact, stream retrieve) for agent-direct mode.
›Adds model support for gpt-5.3-codex and gpt-5.3-chat-latest.
›Raises the default context window from 32k to 128k tokens.
›Raises CORE_MEMORY_BLOCK_CHAR_LIMIT default from 20k to 100k characters.
›Adds effort="max" support in Anthropic model settings where the model supports it.
+4 moreshow less
›Increases the Gemini request timeout default to 600s.
›Conversation creation now compiles and persists a system message immediately at creation time, capturing current memory state.
›Maintains backwards compatibility for the deprecated conversation_id=agent-* path.
›Skills sync in the memory filesystem now maps only skills/{name}/SKILL.md to skills/{name} block labels; other markdown under skills/ is intentionally ignored.
└──▷ BREAKING ON UPGRADE
!The default context window has changed from 32k to 128k — any configuration relying on the old 32k default will now use 128k instead.
!CORE_MEMORY_BLOCK_CHAR_LIMIT default has changed from 20k to 100k — deployments that expected the 20k cap may now allow significantly larger memory blocks.
!Git-backed memory frontmatter no longer emits the limit key; legacy limit keys are removed on merge — any tooling that reads or depends on limit in memory frontmatter will stop seeing it.
!Skills sync now maps ONLY skills/{name}/SKILL.md to block labels; other markdown files under skills/ are ignored — any skills content previously synced via other filenames will no longer appear as blocks.
Semantic Kernel .NET adds SQL Server approximate vector search, DateTime type support across MEVD providers, and PostgreSQL pgvector auto-install.
└──▷ GET THIS VERSION
$ git clone --branch dotnet-1.73.0 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:$ git checkout dotnet-1.73.0
›Adds approximate vector search support for SQL Server in the MEVD (Memory and Embedding Vector Database) layer.
›Adds support for DateTime, DateTimeOffset, DateOnly, and TimeOnly types across MEVD providers.
›Handles automatic installation of pgvector extension on PostgreSQL for MEVD setups.
›Makes PostgreSQL schema nullable in MEVD, enabling more flexible collection definitions.
›Adds validation for mismatched collection keys in MEVD providers to catch configuration errors early.
+1 moreshow less
›Updates Microsoft.Extensions.AI dependencies to 10.3.0 and OpenAI SDK to 2.8.0.
└──▷ BREAKING ON UPGRADE
!SupportsMultipleKeys has been removed from the MEVD layer.
Semantic Kernel vector data adds SQL Server approximate vector search, DateTime type support, and pgvector auto-install for PostgreSQL.
└──▷ GET THIS VERSION
$ git clone --branch vectordata-dotnet-10.0.1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:$ git checkout vectordata-dotnet-10.0.1
›Adds approximate vector search support for SQL Server in the MEVD provider (Microsoft.SemanticKernel.Connectors.SqlServer).
›Supports DateTime, DateTimeOffset, DateOnly, and TimeOnly types across MEVD providers.
›Handles automatic installation of the pgvector extension on PostgreSQL during collection setup.
›Makes the PostgreSQL schema parameter nullable in MEVD, allowing schema-less configurations.
›Adds validation for mismatched collection key types in MEVD providers.
+1 moreshow less
›Deduplicates embedding generation management across MEVD providers.
└──▷ BREAKING ON UPGRADE
!Removes SupportsMultipleKeys from MEVD providers — code referencing this property will fail to compile.
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 v2.1.68 adds 'ultrathink' keyword for on-demand high-effort reasoning and defaults Opus 4.6 to medium effort.
└──▷ GET THIS VERSION
$ git clone --branch v2.1.68 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:$ git checkout v2.1.68
└──▷ TRY IT
Force maximum reasoning depth on a single turn when tackling a complex architectural or security analysis problem.
$ ultrathink — type this keyword in your message to Claude Code to elevate the next turn to high effort
›Re-introduces the ultrathink keyword to trigger high-effort reasoning mode for the next turn.
›Opus 4.6 now defaults to medium effort for Max and Team subscribers, balancing speed and thoroughness.
└──▷ BREAKING ON UPGRADE
!Opus 4 and Opus 4.1 are removed from Claude Code on the first-party API; users with either model pinned are automatically migrated to Opus 4.6.
Pi v0.56.0 adds OpenCode Go provider support, a new branch-summary skip setting, and a Gemini flash-lite fallback model.
└──▷ GET THIS VERSION
$ git clone --branch v0.56.0 https://github.com/earendil-works/pi.git
# already have the repo? check out this version:$ git checkout v0.56.0
›Adds OpenCode Go provider support with opencode-go model defaults and OPENCODE_API_KEY environment variable.
›New branchSummary.skipPrompt setting suppresses branch summarization prompts during tree navigation.
›Adds gemini-3.1-flash-lite-preview as a fallback model in Google provider catalogs when upstream metadata lags.
└──▷ BREAKING ON UPGRADE
!Scoped model entries without an explicit :<thinking> suffix now inherit the current session thinking level when selected, instead of applying a startup-captured default.
!Node OAuth runtime exports are removed from the top-level @mariozechner/pi-ai entry; OAuth login and refresh must now be imported from @mariozechner/pi-ai/oauth.
Zed v0.226.4 adds self-hosted OpenAI-compatible edit predictions, branch diff agent review, and diagnostic badges in the project panel.
└──▷ GET THIS VERSION
$ git clone --branch v0.226.4 https://github.com/zed-industries/zed.git
# already have the repo? check out this version:$ git checkout v0.226.4
└──▷ HOW TO FIND IT
Send the entire branch diff to an AI agent for a one-click code review without leaving the editor.
📍Open the Command Palette and run git: branch diff, then click the Review Diff button that appears in the branch diff view header to send the full diff to the agent panel.
›Adds diagnostic_badges project panel setting (disabled by default) to display error and warning counts next to file names.
›Adds debug_panel::Toggle action to show/hide the debug panel, mirroring terminal_panel::Toggle.
›Adds workspace::ActivateLastPane action for keybindings that focus the last active pane.
›Adds zed: record etw profile action on Windows to collect performance profiles.
›Adds agent::addSelectionToThread (cmd->) behavior change: uses the current cursor line when there is no selection.
+14 moreshow less
›Supports self-hosted OpenAI-compatible servers as a backend for edit predictions.
›Adds a Review Diff button in the git: branch diff view to send the full diff to an agent for review.
›Adds the ability to undo the reject all action when reviewing agent-made changes.
›Adds Claude Sonnet 4.6 support for Amazon Bedrock in the agent.
›Adds diff stat numbers to the git: branch diff view.
›Adds WSL and SSH remote kernel support for the REPL, including spawn/kill kernel protocol messages.
›Adds the ability to clear REPL outputs via action.
›Adds basic HTML rendering in REPL outputs.
›Adds syntax highlighting for HTML character references (·, ', {, etc.) in TSX, JavaScript, and HTML files.
›Adds Go definition highlights for functions, methods, and types.
›Adds .mdc as a recognized Markdown file extension.
›Adds a copy button for project diagnostics messages.
›Improves zed: open performance profiler action to display profiling data from the remote server.
›Improves multi-cursor paste to distribute each clipboard line to its corresponding cursor when line count matches.
Phoenix 13.8.0 adds brute-force login protection, Vercel AI SDK streaming support, and live playground evaluation metrics.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v13.8.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-v13.8.0
›Adds brute-force login protection to the authentication layer, limiting repeated failed login attempts.
›Supports the Pydantic Vercel data stream protocol for agent tracing, enabling observability over Vercel AI SDK streaming responses.
›Playground evaluation metrics now update incrementally as results arrive, rather than waiting for full completion.
›Adds integration icons for Claude and LangGraph in the UI.
›Introduces a file dropzone component in the UI for file uploads.
Composio CLI 0.1.35 adds checksums, cross-compilation, and a simplified install process.
└──▷ GET THIS VERSION
$ git clone --branch @composio/[email protected] https://github.com/ComposioHQ/composio.git
# already have the repo? check out this version:$ git checkout @composio/[email protected]
›Adds checksums, cross-compilation support, and a simplified install flow to the CLI distribution.
THE AI TOOLCHAIN