Heads up This site is currently under heavy development.
Subscribe Get it delivered — the daily firehose, filtered to the tools you run, plus the documentation changes vendors never announce. Compare plans →

The AI Toolchain — issue -239, December 21, 2025

THE AI TOOLCHAIN NO. -239
Tail
THE DAILY RELEASE FIREHOSE
PUBLISHED DECEMBER 21, 2025 · EVERY WEEKDAY
EDITIONS tail grep head diff uniq

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.

VIEW
ISSUE VIEW full issue
Do you prefer this view?
$ tct list   # 5 tools matched
AI & LLM Tooling
◆  AI Coding Agents

OpenAI Codex CLI

Sources Release notes → rust-v0.77.0 NOTES

Lightweight coding agent that runs in your terminal

Codex CLI v0.77.0 adds sandbox mode constraints, normalized TUI scrolling with config knobs, and flag-free MCP OAuth login.

└──▷ GET THIS VERSION
$ git clone --branch rust-v0.77.0 https://github.com/openai/codex.git
# already have the repo? check out this version:
$ git checkout rust-v0.77.0
└──▷ USE IT
Lock a repo to only allow network-disabled sandbox modes, preventing contributors from accidentally running Codex with broader access.
toml
# In your project's requirements.toml
allowed_sandbox_modes = ["workspace-write"]
  • Adds allowed_sandbox_modes field to requirements.toml to restrict which sandbox modes a project permits.
  • Adds tui.scroll_* config settings and normalizes mouse wheel and trackpad scrolling behavior across terminals in TUI2.
  • MCP OAuth login for streamable HTTP servers no longer requires the rmcp_client feature flag — works out of the box.
  • Improves fuzzy file search display consistency by centralizing file-name derivation in codex-file-search.
  • Updates bundled model metadata (models.json) with latest available models.
Was this useful?

SST OpenCode

Sources Release notes → v1.0.185 2 RELEASES · 2025-12-21 NOTES STABLE

The open source coding agent.

OpenCode v1.0.185 adds .ets syntax highlighting, schedule event support, and official MCP SDK integration.

└──▷ GET THIS VERSION
$ git clone --branch v1.0.185 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.0.185
  • Adds syntax highlighting for .ets files.
  • Supports schedule events in GitHub Actions integration.
  • Switches to the official MCP SDK for improved tool schema handling.
1 more release in this issue · 2025-12-21
v1.0.181 NOTES STABLE

OpenCode v1.0.181 adds cursor theme, Catppuccin Frappé theme, default_agent config, and raw markdown docs view.

└──▷ GET THIS VERSION
$ git clone --branch v1.0.181 https://github.com/sst/opencode.git
# already have the repo? check out this version:
$ git checkout v1.0.181
└──▷ USE IT
Set a default agent so every session automatically uses it without specifying one each time.
yaml
# In your opencode config file:
default_agent: "<your-agent-name>"
  • Adds a new Cursor-inspired UI theme via the cursor theme option.
  • Adds Catppuccin Frappé as a built-in color theme.
  • Supports configuring a default_agent across all API and user surfaces.
  • Appending .md to docs page URLs now serves raw markdown content.
Was this useful?

Earendil Works Pi

Sources Release notes → v0.25.3 NOTES

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

Pi v0.25.3 adds Gemini 3 preview models, external editor support, process suspension, and granular skills filtering.

└──▷ GET THIS VERSION
$ git clone --branch v0.25.3 https://github.com/earendil-works/pi.git
# already have the repo? check out this version:
$ git checkout v0.25.3
└──▷ TRY IT
Filter to only skills matching a pattern at startup, e.g. to load only recon-related skills in a focused engagement.
$ pi --skills 'recon*'
Draft a long, structured prompt in your preferred editor rather than the inline input, useful for complex multi-step instructions.
$ # While in the pi prompt, press Ctrl+G to open $EDITOR, write your message, save and quit to send it.
  • Adds gemini-3-pro-preview and gemini-3-flash-preview models to the google-gemini-cli provider.
  • New Ctrl+G keybinding opens the current message in $VISUAL or $EDITOR for external editing.
  • New Ctrl+Z keybinding suspends Pi to the shell; resume with fg.
  • Adds granular skill-source toggles: enableCodexUser, enableClaudeUser, enableClaudeProject, enablePiUser, enablePiProject, plus customDirectories and ignoredSkills settings.
  • New --skills <patterns> CLI flag filters loaded skills by glob pattern; includeSkills setting and glob support for ignoredSkills also added.
Was this useful?
◆  AI Agent Frameworks

OpenClaw

Sources Release notes → v2.0.0-beta2 NOTES

Your own personal AI assistant. Any OS.

OpenClaw v2.0.0-beta2 adds a bundled gateway daemon, a skills platform, WhatsApp QR login, and a browser-control MCP dispatch layer.

└──▷ GET THIS VERSION
$ git clone --branch v2.0.0-beta2 https://github.com/openclaw/openclaw.git
# already have the repo? check out this version:
$ git checkout v2.0.0-beta2
  • New gateway-daemon command packages and runs the gateway as a standalone bundled process with launchd and DMG/zip distribution support.
  • Introduces a managed Skills platform with install metadata, uv-based installers, skill search, and a skills website.
  • Adds models.list API to the Gateway for dynamic model enumeration from connected providers.
  • Adds provider status events and expanded RPC coverage to the Gateway for real-time provider health visibility.
  • Supports Tailscale auth and PAM authentication on the Gateway.
+15 moreshow less
  • Adds bind-mode configuration to the Gateway for flexible network binding.
  • New WhatsApp QR login flow including a dedicated WhatsApp QR login tool in the agent UI.
  • Adds self-chat mode handling for WhatsApp Web.
  • Normalizes Telegram chat IDs with clearer error reporting.
  • Canvas host is now served directly on the Gateway port, eliminating the separate canvasHost port.
  • Adds MCP tool dispatch and native actions to the browser control API.
  • Browser control server can now start without a Playwright dependency present.
  • New agent steering queue for directing in-flight agent behavior.
  • Tool summary streaming and a runtime toggle added to the web/agent UX.
  • Voice wake events now route to the main session.
  • Workspace bootstrap ritual added for new session initialization.
  • Connections settings panel in the macOS app shows provider status and supports QR login.
  • macOS Skills settings redesigned with configurable install targets.
  • Models list in the macOS app is now loaded dynamically from the Gateway.
  • Enriched agent WebSocket logs on the Gateway for improved observability.
└──▷ BREAKING ON UPGRADE
  • !The standalone canvasHost port config has been removed; the canvas host is now served on the Gateway port.
Was this useful?
◆  Local LLM Runtimes

llama.cpp

Sources Release notes → b7502 2 RELEASES · 2025-12-21 NOTES STABLE

Vulkan backend gains async tensor loading with GGML_VK_DISABLE_HOST_VISIBLE_VIDMEM for faster model load times

└──▷ GET THIS VERSION
$ git clone --branch b7502 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b7502
  • Implements set_tensor_async and event interfaces for the Vulkan backend, enabling async load code paths in llama_model_loader::load_all_data.
1 more release in this issue · 2025-12-21
b7492 NOTES STABLE

llama.cpp server gains auto-sleep after configurable idle timeout to free resources automatically

└──▷ GET THIS VERSION
$ git clone --branch b7492 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:
$ git checkout b7492
  • Adds auto-sleep to the llama-server that suspends the server context and frees model resources after a configurable number of idle seconds, reducing memory and GPU pressure when the server is not actively serving requests.
Was this useful?
my-toolchain — 0 tools
paste an install list to detect your tools

A brew list, a Brewfile, requirements.txt, a Dockerfile — or just the product names, free-form. Nothing leaves your browser.

    browse all tools →