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.
AutoGPT v0.6.44 adds Claude Code sandbox execution, external Agent Generator integration, and PostHog/OpenRouter observability for chat.
└──▷ GET THIS VERSION
$ git clone --branch v0.6.44 https://github.com/Significant-Gravitas/AutoGPT.git
# already have the repo? check out this version:$ git checkout v0.6.44
›Adds ClaudeCodeBlock for executing tasks via Claude Code inside an E2B sandbox environment.
›Integrates an external Agent Generator service for automated agent creation workflows.
›Adds PostHog analytics and OpenRouter tracing to the chat system for observability.
›Adds a clarification-questions UI during agent generation so users can refine agent intent before launch.
Framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI empowers agents to work together seamlessly, tackling complex tasks.
CrewAI 1.9.1 adds before/after tool call hooks and structured output support across providers
└──▷ GET THIS VERSION
$ git clone --branch 1.9.1 https://github.com/crewAIInc/crewAI.git
# already have the repo? check out this version:$ git checkout 1.9.1
›Adds before and after tool call hooks in CrewAgentExecutor for intercepting and acting on tool invocations.
›Supports structured outputs and response_format parameter across LLM providers for typed agent responses.
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 v2.1.21 adds Japanese IME input support and automatic Python venv activation in VS Code.
└──▷ GET THIS VERSION
$ git clone --branch v2.1.21 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:$ git checkout v2.1.21
└──▷ USE IT
Opt out of automatic Python venv activation in VS Code if you manage your interpreter another way.
json
// In VS Code settings.json
{ "claudeCode.usePythonEnvironment": false }
›Supports full-width (zenkaku) number input from Japanese IME in option selection prompts.
›Adds automatic Python virtual environment activation in the VS Code extension, ensuring python and pip use the correct interpreter (configurable via claudeCode.usePythonEnvironment).
›Improves read/search progress indicators to show 'Reading…' while in progress and 'Read' when complete.
›Improves Claude's preference for file operation tools (Read, Edit, Write) over bash equivalents (cat, sed, awk).
Stop Zed from inheriting EditorConfig rules from directories above the project root.
ini
root = true
[*]
indent_style = space
indent_size = 2
›Adds "remote": true property in the MCP server settings entry to run MCP servers on the remote server when using remote development.
›Adds use_podman setting to enable Podman support for dev container actions.
›Adds root = true support in .editorconfig files to stop inheriting EditorConfig settings from parent directories; Zed now traverses parent directories to find and apply EditorConfig settings.
›Adds pane: go to older tag and pane: go to newer tag commands for Vim-style tag stack navigation, tracking 'Go to Definition' jumps in a dedicated stack.
›Adds projects: initialize dev container modal to initialize a dev container definition in the current project.
+13 moreshow less
›Adds @diagnostics context to 'New Thread' conversations, letting you feed current language-server diagnostics directly into the Agent Panel.
›Adds ability to install new Agents via the ACP Registry from the Agent panel.
›Adds ability to edit queued agent messages and interrupt the agent while messages are queued.
›Makes prompt queuing the default behavior when sending messages during an ongoing agent generation.
›Adds reasoning content support for the OpenAI provider in the Agent panel.
›Adds a right-click menu item in the agent panel to copy a given agent's response.
›Adds support for viewing and updating project settings files in remote projects via the Settings UI.
›Adds the list of recent projects to the multi-project title bar menu.
›Adds the language server's path to a tooltip shown when hovering on its status in the language server menu.
›Improves pane::CloseActiveItem to close center pane items even when focus is on a dock panel such as the project panel or outline panel.
›Adds support for using a venv-installed ty binary for Python/Ty.
›Opens bundled files, keymap, local release notes, and the log file view within remote windows instead of opening a new local Zed window.
›Opens settings files, keymap files, task files, and debug files within WSL windows instead of opening a new local Zed window.
An open-source AI agent that brings the power of Gemini directly into your terminal.
gemini-cli v0.26.0 adds skill-creator, plan approval mode, agent enable/disable, and undeprecates --prompt
└──▷ GET THIS VERSION
$ git clone --branch v0.26.0 https://github.com/google-gemini/gemini-cli.git
# already have the repo? check out this version:$ git checkout v0.26.0
└──▷ TRY IT
Pass a one-shot prompt directly on the command line without entering the interactive REPL.
$ gemini --prompt "Summarize the last 10 git commits in this repo"
›Introduces skill-creator built-in skill and CJS management tools for authoring and managing agent skills.
›Adds experimental --plan flag and 'plan' approval mode that enforces a strict read-only policy and halts execution on violations before any writes occur.
›Enables and disables individual agents from the CLI.
›Undeprecates the --prompt flag for passing prompts directly on the command line.
›Adds auto alias for default model selection in config.
+14 moreshow less
›Enables agent skills by default.
›Adds conflict detection and warnings when skills override one another.
›Adds security consent prompts when installing skills.
›Adds a Rewind Confirmation dialog and Rewind Viewer component for reviewing conversation history.
›Adds source tracking to policy rules for auditability.
›Adds a generalist agent to the core agent roster.
›Adds admin controls polling and restart prompt for centralized admin management.
›Renames negative settings to positive naming convention (disable* → enable*).
›Adds support for running available commands before MCP servers finish loading.
›Steers the outer agent to delegate to expert sub-agents when they are present.
›Adds experimental extension config setting.
›Adds types and SchedulerStateManager for an event-driven scheduler with reactive tool state.
›Optimizes text buffer and syntax highlighting for large inputs.
›Removes read_many_files from plan approval mode policies so file reads are always permitted without approval.
└──▷ BREAKING ON UPGRADE
!Settings previously named with a disable* prefix are renamed to enable* (e.g., a setting like disableX becomes enableX); existing configs using the old names will need to be updated.
shell-gpt 1.5.0 lets you edit AI-generated shell commands before execution and drops the instructor dependency.
└──▷ GET THIS VERSION
$ git clone --branch 1.5.0 https://github.com/TheR1D/shell_gpt.git
# already have the repo? check out this version:$ git checkout 1.5.0
└──▷ TRY IT
Restore function-calling support after upgrading, required because the instructor package was removed.
$ sgpt --install-functions
›Adds ability to modify a generated shell command before running it, letting you tweak AI output in-place rather than accepting or rejecting wholesale.
└──▷ BREAKING ON UPGRADE
!The instructor package has been removed from dependencies; users who relied on function calling must re-run sgpt --install-functions to reinstall default functions and manually update any custom functions.
›Adds --spec-type CLI flag to enable self-speculative decoding without a separate draft model, using an n-gram map to generate draft tokens from the primary model itself.
›Introduces common_speculative_type enum and a vector of speculative states to support multiple speculation modes within a single server session.
›Adds common_speculative_begin() and common_speculative_state::accept() to the speculative decoding API, enabling structured draft/accept lifecycle management in code integrating llama.cpp.
└──▷ BREAKING ON UPGRADE
!The speculative.draftless_type config field is renamed to speculative.type.
!The --spec-draftless flag is replaced by --spec-type; any existing invocation using --spec-draftless will break.
!The --spec-config flag is removed entirely.
!common_speculative_params is removed from the API.
WebGPU backend gains new unary operators (SOFTPLUS, EXPM1, FLOOR, CEIL, ROUND, TRUNC) and per-thread state isolation.
└──▷ GET THIS VERSION
$ git clone --branch b7851 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:$ git checkout b7851
›Adds SOFTPLUS, EXPM1, FLOOR, CEIL, ROUND, and TRUNC unary operators to the WebGPU backend, each with f16/f32 and inplace/non-inplace shader variants.
›Splits webgpu_context shared state into global state (Instance, Device, Adapter, capabilities) and per-thread state (Queue, pipelines, parameter buffers), enabling safe multi-threaded WebGPU execution.
›Re-enables XIELU unary operator support in the WebGPU backend after correcting IEEE 754 bit-pattern preservation in parameter passing.
Arize Phoenix 12.33.0 adds EMAIL_ATTRIBUTE_PATH for configurable email extraction in OAuth2 flows.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v12.33.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-v12.33.0
└──▷ TRY IT
Map a non-standard email claim path from your IdP's token payload so Phoenix can identify users during OAuth2 login.
$ EMAIL_ATTRIBUTE_PATH=user.contact.email
›Adds EMAIL_ATTRIBUTE_PATH environment variable to OAuth2 configuration, enabling customizable extraction of the email field from identity provider token responses.
THE AI TOOLCHAIN