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 v2.1.217 adds emoji autocomplete, subagent concurrency controls, and new transcript-failure warnings.
└──▷ GET THIS VERSION
$ git clone --branch v2.1.217 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:$ git checkout v2.1.217
└──▷ TRY IT
Allow subagents to spawn one level of nested subagents in an orchestration workflow that relies on hierarchical delegation.
$ CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH=1 claude
›Adds emojiCompletionEnabled setting to control emoji shortcode autocomplete in the prompt input — type :heart: to insert or :hea for suggestions; set emojiCompletionEnabled to false to disable.
›Adds CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS environment variable to cap concurrently-running subagents (default 20), preventing a single message from fanning out unbounded background agents.
›Adds CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH environment variable to allow nested subagent spawning beyond the new default of zero (subagents no longer spawn nested subagents by default).
›Adds FORCE_HYPERLINK=0 environment variable to opt out of always-on footer PR badge hyperlinks, which are now clickable even when terminal hyperlink support cannot be detected (e.g. over SSH/tmux).
›Adds warnings when transcript writes are failing (e.g. disk full) or when session saving is disabled due to an inherited environment variable, rather than silently losing transcripts.
└──▷ BREAKING ON UPGRADE
!Subagents no longer spawn nested subagents by default; set CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH to restore deeper nesting.
$ git clone --branch rust-v0.145.0 https://github.com/openai/codex.git
# already have the repo? check out this version:$ git checkout rust-v0.145.0
›Adds GPT-5.6 Sol as the default Amazon Bedrock model, with experimental managed Bedrock login, custom endpoint, and custom authentication support.
›Expands /import to migrate settings, MCP servers, plugins, sessions, commands, and project-scoped memories from Cursor and Claude Code.
›Adds experimental paginated thread history with efficient resume, search, persisted names, sub-agent support, and memories backed by a dedicated SQLite store.
›Adds audio inputs and tool outputs supporting common local audio formats, and introduces streaming realtime V3 conversations.
›Adds secure, clickable inline visualization links in the terminal UI.
Codex CLI adds route-aware HTTP client pooling, system proxy support across all subsystems, and configurable external session detection limits.
└──▷ GET THIS VERSION
$ git clone --branch rust-v0.145.0-alpha.29 https://github.com/openai/codex.git
# already have the repo? check out this version:$ git checkout rust-v0.145.0-alpha.29
└──▷ TRY IT
Force the Codex installer to use GitHub Releases instead of releases.openai.com when behind a network that blocks the default source.
$ curl -fsSL https://chatgpt.com/codex/install.sh | CODEX_INSTALLER_USE_RELEASES_OPENAI_COM=false sh
›Adds maxSessionAgeDays and maxSessions fields to the external-agent config detection request, making session discovery limits configurable (defaults remain 30-day max age and 50 sessions).
›Adds CARGO_HTTP_CAINFO to the curated custom CA environment variables so Cargo inherits the managed MITM trust bundle in proxy environments.
›Adds an optional opaque providerId field to externalAgentConfig/import for analytics attribution of external agent imports, independently of the migrationSource selector.
›Introduces a route-aware HTTP client pool (RouteAwareClientPool) that resolves the outbound proxy route per request URL, with a bounded cache of 16 routes — enabling correct PAC and system-proxy handling across redirects.
›Extends system proxy settings to cover the daemon updater, remote plugins, plugin startup sync, and cloud environment discovery.
+4 moreshow less
›Preserves thread settings for goal-first and forked threads on restart, so resumed threads use the configuration active when they were created rather than current config.
›Adds an optional releases.openai.com installer source, with CODEX_INSTALLER_USE_RELEASES_OPENAI_COM controlling fallback to GitHub Releases.
›Mirrors Rust release artifacts and metadata to Cloudflare R2 as an additional distribution channel.
›Supports alpha hotfix release versions in the version scheme (mapping Python aN.postM to -alpha.N.M tags) across installers, npm publishing, and release validation.
$ git clone --branch rust-v0.145.0-alpha.28 https://github.com/openai/codex.git
# already have the repo? check out this version:$ git checkout rust-v0.145.0-alpha.28
›Supports custom transports for Amazon Bedrock, enabling non-default network or proxy configurations when connecting to Bedrock-backed models.
›Adds an app-server API for reading installed app runtime state.
›Adds audio input and output support: audio variants in user input protocols, audio inputs forwarded to the Responses API, audio output in dynamic tools and code mode, and audio history gated by model input modalities.
›Adds SessionEnd hooks for thread teardown, enabling cleanup logic when a session ends.
›Adds batched executor capability discovery.
+18 moreshow less
›Adds occurrence search for paginated threads.
›Renders inline visualization links in the TUI.
›Renders streamed Markdown incrementally in the TUI, reducing layout work and redundant redraws while streaming.
›Allows users to remember the working directory for resumed sessions.
›Allows publishing plugins through share updates.
›Allows selecting path-backed agents in the TUI agent picker.
›Confirms usage-limit resets before redemption, preventing accidental redemption.
›Suppresses empty multi-agent mode messages in the TUI.
›Reports detailed session import error types.
›Tracks collaboration mode instructions and permission instructions in world state.
›Centralizes thread MCP connections in McpRuntime.
›Supports ChatGPT-branded Desktop app builds.
›Adds bounded batch lookups and batched persistent history reads for message history, improving reverse-search performance.
›Launches managed network proxies on remote executors.
›Makes parent-owned sub-agent threads read-only in the TUI.
Codex CLI alpha.27 lets Esc interrupt a running session and updates bundled model metadata for GPT-5.6 and GPT-5.5.
└──▷ GET THIS VERSION
$ git clone --branch rust-v0.145.0-alpha.27 https://github.com/openai/codex.git
# already have the repo? check out this version:$ git checkout rust-v0.145.0-alpha.27
›Allows pressing Esc to interrupt the current agent session while a task is running, with Esc now documented as an interrupt key in the status indicator.
›Updates bundled GPT-5.6 model variant context windows to 272,000 tokens and refreshes model instructions, message configuration, and reasoning-summary support across the model catalog.
›Adds personality instruction variables for GPT-5.6 model variants and adds a GPT-5.5 availability notice to the bundled catalog.
›Runs write_stdin in parallel across unified-exec sessions, improving throughput for multi-session workloads.
└──▷ BREAKING ON UPGRADE
!The auto_review and permissions message fields and the legacy supports_reasoning_summary_parameter flag are removed from the bundled model catalog.
$ git clone --branch cloud-1.47.0 https://github.com/All-Hands-AI/OpenHands.git
# already have the repo? check out this version:$ git checkout cloud-1.47.0
$ git clone --branch cloud-1.47.0 https://github.com/OpenHands/OpenHands.git
# already have the repo? check out this version:$ git checkout cloud-1.47.0
›Agent Canvas is now protected behind SaaS authentication, restricting access to authenticated users only.
›Adds /login command for llama.cpp router connection setup and /llama subcommand for Hugging Face model search, download, explicit load, unload, and live progress tracking.
›Adds get_available_thinking_levels RPC command and RpcClient.getAvailableThinkingLevels() method for querying model thinking-level options.
›Adds full provider extension registration, allowing extensions to supply complete pi-ai providers with native authentication, model refresh, filtering, and custom streaming behavior.
›Adds Qwen Token Plan and Qwen Token Plan China as built-in providers with regional endpoints and API-key authentication.
›Exports message and tool execution lifecycle event types from the package root for downstream consumers.
+1 moreshow less
›Expands usage accounting to persist and include tool, compaction, and branch-summary usage in session totals, footer totals, and session statistics.
langchain-anthropic 1.5.0 adds reasoning_effort as a standard chat model parameter.
└──▷ GET THIS VERSION
$ git clone --branch langchain-anthropic==1.5.0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:$ git checkout langchain-anthropic==1.5.0
└──▷ USE IT
Tune how much reasoning an Anthropic model applies before responding — useful when balancing latency against answer quality.
python
from langchain_anthropic import ChatAnthropic
llm = ChatAnthropic(model="claude-sonnet-4-5", reasoning_effort="low")
response = llm.invoke("Explain the RSA algorithm.")
print(response.content)
›Adds reasoning_effort as a standard chat model parameter for controlling reasoning depth in Anthropic models.
langchain-fireworks 1.5.0 adds reasoning_effort as a standard chat model parameter.
└──▷ GET THIS VERSION
$ git clone --branch langchain-fireworks==1.5.0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:$ git checkout langchain-fireworks==1.5.0
›Adds reasoning_effort as a standard chat model parameter for Fireworks-hosted models.
langchain-xai 1.3.0 adds reasoning_effort as a standard chat model parameter and XAI_API_BASE/base_url support.
└──▷ GET THIS VERSION
$ git clone --branch langchain-xai==1.3.0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:$ git checkout langchain-xai==1.3.0
└──▷ USE IT
Control reasoning depth on a per-call basis when working with xAI models that support tiered reasoning.
python
from langchain_xai import ChatXAI
llm = ChatXAI(model="grok-3-mini", reasoning_effort="high")
response = llm.invoke("Explain the MITRE ATT&CK framework in detail.")
print(response.content)
Point langchain-xai at a proxy or self-hosted xAI-compatible endpoint without modifying source code.
LangChain OpenAI 1.4.0 adds reasoning_effort as a standard chat model parameter.
└──▷ GET THIS VERSION
$ git clone --branch langchain-openai==1.4.0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:$ git checkout langchain-openai==1.4.0
└──▷ USE IT
Control reasoning depth on an OpenAI model invocation to balance latency against thoroughness.
python
from langchain_openai import ChatOpenAI
llm = ChatOpenAI(model="o3", reasoning_effort="low")
response = llm.invoke("Explain the threat model for a zero-trust architecture.")
print(response.content)
›Adds reasoning_effort as a standard chat model parameter for OpenAI chat models.
langchain-core 1.5.0 adds reasoning_effort as a standard chat model parameter.
└──▷ GET THIS VERSION
$ git clone --branch langchain-core==1.5.0 https://github.com/langchain-ai/langchain.git
# already have the repo? check out this version:$ git checkout langchain-core==1.5.0
›Adds reasoning_effort as a standard chat model parameter, enabling portable control of model reasoning depth across chat model providers.
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.
emisar v0.33.0 adds uninstall/reset runner identity flags, trust-gated action dispatch, and a new incident-response skill.
└──▷ GET THIS VERSION
$ git clone --branch v0.33.0 https://github.com/AndrewDryga/emisar.git
# already have the repo? check out this version:$ git checkout v0.33.0
└──▷ TRY IT
Cleanly remove a decommissioned runner's identity and token without wiping its logs or config, so the runner can be re-enrolled fresh.
$ install.sh --uninstall
Reset a runner's external identity in an automated pipeline when rotating enrollment keys, skipping the interactive prompt.
$ install.sh --reset-identity
›Adds --uninstall flag to install.sh to remove the cached token, legacy token file, and generated runner identity while preserving configuration, local evidence, and logs.
›Adds --reset-identity flag to install.sh for unattended runner identity resets when supplying a different enrollment key.
›Adds --purge flag to install.sh that extends --uninstall by also removing the retained configuration, evidence, and log files.
›MCP catalog, search, exact lookup, runner inspection, and runbook recovery now filter out pending, rejected, revoked, retired, hash-mismatched, and incomplete pack versions from model-visible results.
›Runner UI locks the Run button and surfaces a diagnostic explanation when an advertised action's pack is untrusted or retired; dispatch revalidates the contract so stale pages or tool calls cannot execute after trust changes.
+5 moreshow less
›Runbook recovery now fails closed if trust changes between action inspection and execution, preventing hidden pack version leakage through recovery results.
›ChatGPT tool annotations now distinguish read-only calls from mutations, and domain verification accepts OpenAI's text challenge.
›Adds a new public respond-to-production-incidents skill providing customer agents a bounded observe-diagnose-act-verify workflow.
›The Packs page now follows the live catalog and gives unadvertised versions one day to disappear before cleanup.
›OAuth return targets now preserve their original local query string through SSO and registration flows.
Arize Phoenix Evals 3.2.0 adds a user friction evaluator for detecting friction in AI interactions.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-evals-v3.2.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-evals-v3.2.0
›Adds a user friction evaluator to detect and score friction points in AI assistant conversations.
THE AI TOOLCHAIN