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 v2.1.223 adds owner wildcards for marketplace policies, a /teleport hint for cloud sessions, and /review alias improvements.
└──▷ GET THIS VERSION
$ git clone --branch v2.1.223 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:$ git checkout v2.1.223
└──▷ USE IT
Block all marketplace repos under a GitHub org in a single managed-settings entry, rather than listing each repo individually.
json
{
"blockedMarketplaces": ["my-org/*"]
}
Pick up a cloud session on your local machine when the /teleport hint appears in a cloud session.
$ claude --teleport <session id>
Run a deep cloud code review on a specific pull request.
$ /code-review ultra 42
›Adds owner wildcard entries ("owner/*") to strictKnownMarketplaces and blockedMarketplaces managed settings, allowing admins to permit or block all marketplace repos under an entire GitHub org in one entry.
›Changes CLAUDE_CODE_DISABLE_1M_CONTEXT to apply the 200K auto-compaction cap to every Claude model with a native 1M window dynamically rather than a fixed list, and adds a startup warning when auto-compaction is not holding the session to 200K.
›Adds a /teleport hint in cloud sessions showing how to continue work locally with claude --teleport <session id>.
›Makes /review an alias for /code-review, which reviews the current diff or a PR via /code-review <level> <pr#>; /code-review ultra triggers a deep cloud review.
›Changes /code-review with no effort level to reuse the last typed level, so re-running the command does not require re-specifying the effort.
+1 moreshow less
›Adds a warning when a workflow agent, forked skill, slash command, or resumed background agent requests a subagent model that is restricted and the parent model runs instead.
Cline CLI v3.0.51 adds meta/muse-spark-1.2-contributor model and unified reasoning-effort control across all providers.
└──▷ GET THIS VERSION
$ git clone --branch cli-v3.0.51 https://github.com/cline/cline.git
# already have the repo? check out this version:$ git checkout cli-v3.0.51
›Adds meta/muse-spark-1.2-contributor as a selectable model on the Cline provider, alongside a refreshed model catalog.
›Reasoning effort now applies consistently across all providers (including Ollama) without per-provider thinking overrides, and requests to disable reasoning are respected everywhere.
Lightweight coding agent that runs in your terminal
Codex CLI alpha adds thread section ordering, thread/section/move API, named forked chats, and MCP OAuth credential isolation.
└──▷ GET THIS VERSION
$ git clone --branch rust-v0.147.0-alpha.13 https://github.com/openai/codex.git
# already have the repo? check out this version:$ git checkout rust-v0.147.0-alpha.13
›Adds --psp flag to the app-server to enable process-only PSP routing for first-party ChatGPT requests.
›Enables naming forked chats from the TUI.
›Loads cloud-managed servers in MCP CLI commands.
›Isolates MCP OAuth credentials by environment and restricts hosted MCP credentials to local environments.
›Exposes MCP read-only hints in tool call items.
+8 moreshow less
›Distinguishes unknown MCP authentication status as a separate state.
›Enables skills in the MCP server.
›Supports Enterprise automation account plans.
›Denies network access when an allow amendment fails, tightening sandbox enforcement.
›Requires explicit sandbox permissions with shell justifications.
›Restricts shell_command to a single local environment.
›Detects connectors used in external agent sessions.
›Records attempted tool calls in response metadata.
└──▷ BREAKING ON UPGRADE
!Legacy --full-auto handling is removed from codex exec.
!Legacy Linux bundle archives are no longer published.
Pi v0.84.0 adds fullscreen TUI, Mermaid/LaTeX rendering, per-directory context overrides, Baseten provider, and custom sampling params.
└──▷ GET THIS VERSION
$ git clone --branch v0.84.0 https://github.com/earendil-works/pi.git
# already have the repo? check out this version:$ git checkout v0.84.0
└──▷ TRY IT
Launch Pi in fullscreen TUI mode for a distraction-free session with an independently scrollable transcript.
$ pi --tui-mode fullscreen
Override context for a specific subdirectory without affecting parent-directory context files.
$ # Place directory-specific instructions in AGENTS.override.md
cat > src/api/AGENTS.override.md << 'EOF'
# API module context
Always validate inputs with Zod. Do not modify generated files.
EOF
›Adds --tui-mode fullscreen CLI flag and /settings toggle for fullscreen TUI mode with a sticky editor, independently scrollable transcript, and draggable scrollbar configurable via auto, always, or hidden modes.
›Adds AGENTS.override.md per-directory context file that replaces AGENTS.md or CLAUDE.md in the same directory while preserving context from parent directories.
›Adds arbitrary OpenAI-compatible model sampling parameters via samplingParams in models.json, model overrides, extension providers, and stream options.
›Adds opt-in vLLM thinking_token_budget support for OpenAI-compatible models, reserving output tokens for the final answer.
›Adds built-in Baseten provider support authenticated via BASETEN_API_KEY environment variable, with zai-org/GLM-5.2 as the default model.
+10 moreshow less
›Adds AI_AGENT=pi to CLI and RPC child-process environments for generic agent attribution.
›Adds support for OpenAI-compatible streams that omit finish_reason, using compat.supportsFinishReason to infer normal and tool-use stops.
›Adds chainable pi.registerMarkdownTransformer() hooks for display-only transformation of user and assistant Markdown.
›Adds configurable Mermaid diagram rendering as themed Unicode in interactive transcripts, including optional rendering while streaming.
›Adds terminal-friendly Unicode rendering for LaTeX expressions in Markdown transcripts.
›Adds opt-in Ctrl+P/Ctrl+N prompt history navigation in the TUI editor.
›Adds experimental remote-session client APIs: the transport-neutral PiClient, CBOR protocol, Unix-socket transport, and @earendil-works/pi-coding-agent/clientRemoteSession controller with transcript reducers.
›Adds vendor-neutral telemetry contracts with agent-owned typed AI-request and harness schemas, composed span starters, and callback helpers.
›Adds structured Amazon Bedrock failure diagnostics including HTTP status, modeled error code, and AWS request ID.
›Adds CredentialSynchronizationError for credential changes that commit successfully but fail to synchronize local model state.
└──▷ BREAKING ON UPGRADE
!The ModelsStreamTransforms interface is renamed to ModelsRequestTransforms; extensions referencing the old name will break.
!JSON and RPC message_update events now emit only assistantMessageEvent deltas; the cumulative message and assistantMessageEvent.partial fields are removed. Clients must assemble deltas between message_start and message_end.
!ModelRegistry.getApiKeyAndHeaders() now returns ProviderHeaders with string | null values; extensions that inspect returned headers must handle null.
!ModelRegistry.refresh() now accepts ModelsRefreshOptions and returns ModelsRefreshResult instead of discarding cancellation and provider errors.
!ModelRuntime.setRuntimeApiKey() now accepts auth cancellation options instead of catalog refresh options; call refresh({ providers: [providerId], signal }) separately when remote freshness is required.
!Config-form extension OAuth refreshToken(credentials, signal) callbacks are now required to accept and honor a concrete abort signal.
!Dynamic provider refresh context store access is replaced with the read-only context.stored snapshot and generation-checked context.publish() transaction; handwritten Provider.refreshModels() implementations must be migrated.
!The legacy JSONL and in-memory repository APIs are removed; use pi-agent-core's v4 JsonlSessionRepo or InMemorySessionRepo implementing the new SessionRepo contract.
!Custom harness file-system implementations must now provide FileSystem.renameFile() with same-filesystem replacement semantics.
!RemoteSession.sessions no longer exposes runtime phase, model, thinking, attachment, or lock state; this information is now only available from acquired SessionSnapshot values.
!The v2 session and AgentHarness API experimental subpaths are removed; use pi-agent-core's default export instead.
!The Session, SessionStorage, and SessionRepo APIs are replaced with v4 lane-based versions including durable operation records, global facts, shared sequence numbers, and tree-scoped lane views.
Prevent silent CI hangs on long-running /review streams by capping stream lifetime to 10 minutes.
$ QWEN_STREAM_MAX_LIFETIME_MS=600000 qwen review
Bootstrap a repo-specific review context manifest so /review plans include project-aware context.
$ qwen review repo-context
›Adds security.allowedInsecureVoiceBaseUrls setting to route voice transcription through specific HTTP or private-network ASR gateways in managed deployments.
›Introduces .qwen/review-context.json declarative manifest and the review repo-context command to customize review plans with repository-specific context.
›Adds QWEN_STREAM_MAX_LIFETIME_MS setting to cap the total lifetime of streaming responses and prevent silent hangs in GitHub-triggered /review runs.
›Enables rendering inline terminal images from model and tool outputs in the interactive CLI, supporting Kitty, Ghostty, and chafa.
›Adds observability for REST SSE connections with stable UUIDs and telemetry for stream lifecycle events, including slow-client warnings and evictions.
Qwen Code live-host-v0.1.0 adds local GitHub auth, mid-turn read-only shell commands, and reverse-dependency diff testing.
└──▷ GET THIS VERSION
$ git clone --branch live-host-v0.1.0 https://github.com/QwenLM/qwen-code.git
# already have the repo? check out this version:$ git checkout live-host-v0.1.0
›Supports local GitHub (gh) authentication via channels, enabling offline or self-hosted auth flows.
›Enables the web shell to run read-only informational commands immediately mid-turn, without waiting for the current turn to complete.
›Adds reverse-dependency closure testing for diffs in the review workflow, surfacing dependency-impact caveats alongside review results.
Semantic Kernel Python 1.44.1 adds MCP tool approval callback for Azure AI Agent and skips colliding MCP tool names.
└──▷ GET THIS VERSION
$ git clone --branch python-1.44.1 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:$ git checkout python-1.44.1
›Adds MCP tool approval callback for Azure AI Agent, enabling programmatic review and approval of MCP tool invocations before execution.
›Skips MCP tools and prompts whose normalized names collide, preventing silent conflicts when multiple MCP sources expose identically named tools.
›Encodes OpenAPI server variable values to handle special characters in server URL construction.
└──▷ BREAKING ON UPGRADE
!The MCP tool approval callback for Azure AI Agent is marked as a breaking change — existing Azure AI Agent setups using MCP tools may require updates to accommodate the new callback interface.
Semantic Kernel 1.79.0 adds Ollama thinking support and TimeProvider injection for deterministic testing
└──▷ GET THIS VERSION
$ git clone --branch dotnet-1.79.0 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:$ git checkout dotnet-1.79.0
└──▷ USE IT
Enable Ollama's extended thinking/reasoning mode when executing a prompt, useful for tasks requiring chain-of-thought output.
csharp
var settings = new OllamaPromptExecutionSettings { Think = true };
var result = await kernel.InvokePromptAsync(prompt, new(settings));
Inject a fake TimeProvider into TimePlugin so unit tests get deterministic timestamps instead of the real wall clock.
csharp
var fakeTime = new FakeTimeProvider();
fakeTime.SetUtcNow(new DateTimeOffset(2025, 1, 1, 0, 0, 0, TimeSpan.Zero));
var timePlugin = new TimePlugin(fakeTime);
kernel.Plugins.AddFromObject(timePlugin);
›Adds Think property to OllamaPromptExecutionSettings, enabling extended thinking/reasoning mode for Ollama-backed prompts.
›Adds TimeProvider injection to TimePlugin, enabling deterministic time-based testing by supplying a custom or fake clock.
!Upgrades Prompty.Core to 2.0.0-beta.3; the commit is marked [BREAKING] — projects depending on the previous Prompty.Core version may require migration.
THE AI TOOLCHAIN