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.
Agno v2.7.0 adds service account PATs, a full CLI (agnoctl), MCP Interface v2, an eval suite runner, and a discovery endpoint.
└──▷ GET THIS VERSION
$ git clone --branch v2.7.0 https://github.com/agno-agi/agno.git
# already have the repo? check out this version:$ git checkout v2.7.0
└──▷ TRY IT
Wire up all supported coding-agent MCP clients (Claude Code, Cursor, Codex, etc.) to a running AgentOS in one step, minting a named shared PAT.
$ uvx agno connect --name ci-bot
Rotate a compromised PAT without touching your AgentOS deployment.
$ agno tokens revoke agno_pat_abc123xyz
›Adds agno connect command (via uvx agno connect) that discovers an AgentOS, mints per-client PATs (--name for a shared token), writes MCP config for Claude Code, Claude Desktop, Cursor, Codex, and ChatGPT, and verifies each connection — no manual JSON editing.
›Adds agno tokens create/list/revoke commands to mint, inventory, and rotate Personal Access Tokens (agno_pat_... machine identities) with SHA-256-hashed storage, per-user scoping, and revocation support.
›Adds agno create command to scaffold a new AgentOS project from a template (agentos-<provider>).
›Adds agno up, agno down, agno restart, and agno status commands for lifecycle management of local AgentOS deployments.
›New agnoctl CLI distributed on PyPI and invoked as agno — the unified command surface for all of the above.
+6 moreshow less
›New agno.eval package introducing Case and run_cases with a CLI runner supporting team subjects and numeric judge scoring.
›New GET /info discovery endpoint reporting agno_version, mcp.enabled, mcp.path, and auth_mode so external tooling can inspect an AgentOS before connecting.
›New MCP Interface v2 exposes an 8-tool operator surface at /mcp: get_agentos_config, run_agent, run_team, run_workflow, continue_run, cancel_run, get_sessions, and get_session_runs; includes MCP progress notifications for long-running tools and a HITL continue/cancel lifecycle.
›Adds result_mode="full" escape hatch in MCPServerConfig to opt out of the new trimmed run-result shape.
›Single AuthMiddleware on the parent app now covers REST, /mcp, and WebSocket transports; JWTMiddleware is preserved as an alias for backward compatibility.
›A2A/AGUI routes now enforce authorization (scope-mappings merged per-interface at the mount prefix, gating custom prefixes too).
└──▷ BREAKING ON UPGRADE
!MCP surface shrunk from 19 to 8 tools — session-write and memory-CRUD tools are no longer exposed via MCP; MCPServerConfig(include_tags={"memory"}) now fails Pydantic validation and callers must use REST endpoints instead.
!MCP run results are trimmed by default — clients consuming the raw RunOutput shape must handle the new trimmed shape or set result_mode="full" in MCPServerConfig.
!AgentOS(authorization=True) without JWT keys now raises ValueError at construction instead of silently serving an open instance; set JWT_VERIFICATION_KEY or JWT_JWKS_FILE env vars, or pass verification_keys/jwks_file via authorization_config.
!The AGNO_OS_URL environment variable is renamed to AGENTOS_URL; any environment or CI config referencing the old name must be updated.
RealtimeAgent defaults to gpt-realtime-2.1 and SQLAlchemySession gains a Unicode storage option.
└──▷ GET THIS VERSION
$ git clone --branch v0.18.0 https://github.com/openai/openai-agents-python.git
# already have the repo? check out this version:$ git checkout v0.18.0
›Adds Unicode storage option to SQLAlchemySession for broader character-set support in session persistence.
›Changes the default model for RealtimeAgent to gpt-realtime-2.1.
└──▷ BREAKING ON UPGRADE
!The default model for RealtimeAgent is now gpt-realtime-2.1; any existing code that relied on the previous default model will silently switch behaviour on upgrade.
Semantic Kernel Python 1.44.0 hardens MCP tool filtering, OpenAPI server validation, and request handling.
└──▷ GET THIS VERSION
$ git clone --branch python-1.44.0 https://github.com/microsoft/semantic-kernel.git
# already have the repo? check out this version:$ git checkout python-1.44.0
›Enforces excluded_functions on the MCP tool invocation path, ensuring excluded tools cannot be called even when reached through MCP.
›Adds OpenAPI server URL validation to catch misconfigured or untrusted server targets before requests are dispatched.
›Defaults MCP SSE server samples to loopback with host validation, reducing exposure from unintended network interfaces.
›Hardens OpenAPI operation path handling for more consistent operation selection and request targeting.
└──▷ BREAKING ON UPGRADE
!Runtime handling has been updated in a breaking change ([Breaking] Update runtime handling); existing code relying on the previous runtime behaviour may require updates on upgrade to 1.44.0.
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 v2.1.203 adds login-expiry warnings, MCP roots integration, a manual-permission badge, and a VSCode remote-control toggle.
└──▷ GET THIS VERSION
$ git clone --branch v2.1.203 https://github.com/anthropics/claude-code.git
# already have the repo? check out this version:$ git checkout v2.1.203
›Adds a warning when your login session is about to expire so you can re-authenticate before background sessions are interrupted.
›Adds a grey ⏸ badge to the footer when manual permission mode is active, making the current mode always visible.
›Exposes the session's additional working directories via MCP roots/list, with notifications/roots/list_changed fired whenever the set changes.
›Moves 'claude command missing or broken' startup warnings out of the launch flow into /doctor and /status for on-demand review.
›Changes the claude agents detail views (background tasks, diff, workflow) to close with Esc instead of the left arrow key.
+2 moreshow less
›Redesigns the empty claude agents view to always display organized sections (Needs Input / Working / Completed) with descriptions.
›[VSCode] Adds a Settings toggle for 'Enable Remote Control for all sessions'.
└──▷ BREAKING ON UPGRADE
!The left arrow key no longer closes the background tasks, diff, and workflow detail views — press Esc instead.
Autonomous coding agent as an SDK, IDE extension, or CLI assistant.
Cline CLI v3.0.38 adds mode-tinted markdown, smarter read_files input handling, and a 128K fallback context for under-specified models.
└──▷ GET THIS VERSION
$ git clone --branch cli-v3.0.38 https://github.com/cline/cline.git
# already have the repo? check out this version:$ git checkout cli-v3.0.38
›Adds mode-aware markdown tinting: assistant responses are now visually accented by the plan/act mode they were produced in.
›Improves read_files tolerance: malformed line-range entries (start_line/end_line) sent as separate array items by weaker models are automatically coalesced onto the preceding file path instead of being rejected.
›Adds a 128K input-token fallback for catalog models that don't report a context window, giving under-specified models a usable context budget.
›Changes the thinking-level picker default cursor to Medium instead of Off.
Copilot CLI v1.0.69 adds /mcp list, auto allow-all LLM-judged approvals, delegate PR branching, and a /plugins dashboard.
└──▷ GET THIS VERSION
$ git clone --branch v1.0.69 https://github.com/github/copilot-cli.git
# already have the repo? check out this version:$ git checkout v1.0.69
└──▷ TRY IT
List all attached MCP servers and their current status mid-session without interrupting the agent.
$ /mcp list
Keep the CLI in autopilot mode automatically after a task completes, useful for unattended batch workflows.
yaml
stayInAutopilot: true
Create a delegate PR targeting a specific branch instead of the current branch.
$ /delegate --base main
›Adds /mcp list subcommand to show attached MCP servers and their status; /mcp list and /plugin list can now run while the agent is working.
›Adds stayInAutopilot setting (default false) that keeps the CLI in autopilot mode after an autopilot task completes.
›Adds /delegate --base flag to choose a target branch when creating delegate PRs; default target is now the current branch.
›Adds /allow-all auto mode that auto-approves requests an LLM judge evaluates as acceptable; now requires experimental mode (/experimental on or --experimental) instead of the AUTO_APPROVAL env var or feature flag.
›Adds minimal reasoning effort support for the gemini-3.5-flash model.
+23 moreshow less
›Adds a /plugins dashboard to manage installed plugin extensions.
›Supports reloading installed plugin extensions without restarting a session.
›Supports signing in to MCP servers through the CLI OAuth callback flow.
›Enables opening the /mcp manager while the agent is working to enable or disable servers mid-turn; add, edit, delete, and re-auth actions remain paused until the turn finishes.
›Adds file and folder completion to /sandbox path entries.
›web_fetch now follows the active sandbox network policy, denying outbound or local targets the policy blocks, and supports one-time bypass approval via sandbox.allowBypass.
›Displays reasoning-effort labels in the CLI footer.
›Shows descriptions for sandbox userPolicy settings in /settings.
›Rejects explicitly empty --name= values when starting a session.
›Rejects empty --session-id= and --resume= values instead of silently ignoring them.
›Applies --reasoning-effort reliably when the CLI starts.
›Surfaces skill loading errors and warnings in skill list.
›Auto-approval timeline entries now include the request subject such as command, path, URL, or tool name.
›Shows /rubber-duck in pre-auth help and self-documentation.
›Labels built-in file edits with a (sandbox policy) badge instead of (sandboxed), reflecting best-effort sandbox policy behavior.
›Shows disabled servers as disabled in /lsp test.
›Allows WSL UNC paths for local files in the CLI.
›Warns when multiple plugins define the same MCP server.
›Warns when static context uses most of the prompt budget and blocks requests when little conversation room remains.
›Confirms before resuming a remote session from a different repository.
›Delays read-only remote session creation until the first message is sent.
›Includes files inside new untracked directories in /diff local edits.
›Updates a backgrounded session's branch label in the Sessions split view when its working directory changes.
└──▷ BREAKING ON UPGRADE
!/allow-all auto now requires experimental mode (/experimental on or --experimental) instead of the AUTO_APPROVAL env var or feature flag; existing setups relying on the env var or feature flag will need to switch.
!The (sandboxed) badge on built-in file edits is renamed to (sandbox policy).
OpenHands 1.42.0 adds a super-admin management API for granting, revoking, and listing admin privileges in enterprise deployments.
└──▷ GET THIS VERSION
$ git clone --branch cloud-1.42.0 https://github.com/All-Hands-AI/OpenHands.git
# already have the repo? check out this version:$ git checkout cloud-1.42.0
›New enterprise super-admin management endpoint lets administrators programmatically grant, revoke, and list super-admin roles via API.
SYCL backend gains support for cross_entropy_loss and cross_entropy_loss_back operations in llama.cpp b9902.
└──▷ GET THIS VERSION
$ git clone --branch b9902 https://github.com/ggml-org/llama.cpp.git
# already have the repo? check out this version:$ git checkout b9902
›Adds cross_entropy_loss and cross_entropy_loss_back operator support to the SYCL backend, enabling GPU-accelerated training-related operations on Intel SYCL devices.
Phoenix v17.20.0 adds user-selected metric charts above project tables in the UI.
└──▷ GET THIS VERSION
$ git clone --branch arize-phoenix-v17.20.0 https://github.com/Arize-ai/phoenix.git
# already have the repo? check out this version:$ git checkout arize-phoenix-v17.20.0
›Adds user-selected metric charts displayed above project tables on the Projects page, letting practitioners pin and view key metrics at a glance.
THE AI TOOLCHAIN