Daily digest · 2026-10-24 (UTC)
What broke for AI agents on 24 Oct 2026
6 incidents on 3 status pages.
Final. Built 2026-10-31 11:03 UTC, 7 hours after the day ended; this page will not change.
What broke
- Anthropic: “Elevated errors for Claude Fable 5.1, Sonnet 5.5, and Haiku 5.5”critical impact · opened Oct 23 23:37 UTC · resolved 00:11 UTC (34 minutes) · Claude API (api.anthropic.com) | Claude Code | Claude Cowork | claude.aistatus page ↗provider-incidents v58
- Anthropic: “Elevated errors for Claude Fable 5 and Claude Fable 5.1”major impact · opened 05:16 UTC · resolved 05:53 UTC (37 minutes) · Claude API (api.anthropic.com) | Claude Code | Claude Cowork | claude.aistatus page ↗provider-incidents v59
- Anthropic: “Elevated errors for Claude Opus 5.5”major impact · opened 00:31 UTC · resolved 00:37 UTC (6 minutes) · Claude API (api.anthropic.com) | Claude Code | Claude Cowork | claude.aistatus page ↗provider-incidents v58
- Cloudflare: “Increased error rate in Workflow Operations”minor impact · opened 13:13 UTC · resolved 13:57 UTC (44 minutes) · Workflowsstatus page ↗provider-incidents v60
- Cloudflare: “Network Performance Issues in Ashburn”minor impact · opened 01:01 UTC · resolved 03:09 UTC (2 h 8 min) · Artifacts | Ashburn, United States - (IAD) | D1 | Durable Objects | Network | Queues | R…status page ↗provider-incidents v65
- GitHub: “Grok 4.6 and Grok 4.7 Service Issues”minor impact · opened Oct 23 23:41 UTC · resolved 00:23 UTC (42 minutes) · Copilot AI Model Providersstatus page ↗provider-incidents v58
Shipped
No new major version of the SDKs and tools we follow.
16 other releases on the day
- ai 7.0.153 (npm)released 17:57 UTC · previous 7.0.152agent-sdk-releases v79
- anthropics/claude-agent-sdk-python v0.2.174published 00:24 UTC · previous v0.2.173release notes ↗agent-tool-releases v75
- BerriAI/litellm v1.103.8published 08:11 UTC · previous v1.107.0release notes ↗agent-tool-releases v76
- BerriAI/litellm v1.104.6published 08:40 UTC · previous v1.103.8release notes ↗agent-tool-releases v77
- BerriAI/litellm v1.105.4published 08:52 UTC · previous v1.104.6release notes ↗agent-tool-releases v77
- BerriAI/litellm v1.106.3published 09:34 UTC · previous v1.105.4release notes ↗agent-tool-releases v77
- claude-agent-sdk 0.2.174 (pypi)released 00:21 UTC · previous 0.2.173agent-sdk-releases v76
- crewAIInc/crewAI 1.15.30published 01:56 UTC · previous 1.15.29release notes ↗agent-tool-releases v75
- langchain-ai/langgraph 1.2.17published 01:08 UTC · previous 1.2.16release notes ↗agent-tool-releases v75
- litellm 1.106.3 (pypi)released 09:14 UTC · previous 1.107.0agent-sdk-releases v78
- ollama/ollama v0.40.11published 00:30 UTC · previous v0.40.10release notes ↗agent-tool-releases v78
- vercel/ai @ai-sdk/[email protected]published 18:01 UTC · previous @ai-sdk/[email protected]release notes ↗agent-tool-releases v78
- vercel/ai @ai-sdk/[email protected]published 18:01 UTC · previous @ai-sdk/[email protected]release notes ↗agent-tool-releases v78
- vercel/ai @ai-sdk/[email protected]published 18:01 UTC · previous @ai-sdk/[email protected]release notes ↗agent-tool-releases v78
- vercel/ai @ai-sdk/[email protected]published 18:01 UTC · previous @ai-sdk/[email protected]release notes ↗agent-tool-releases v78
- vercel/ai [email protected]published 17:57 UTC · previous [email protected]release notes ↗agent-tool-releases v78
Measured on the day
- 150 round-trips to 23 endpoints agents depend on, 0 failedmodel APIs, registries, chains, the x402 facilitator, status pages and two baselines · measured from macbook-kragent-api-observatory v100
- 447 of 500 remote MCP endpoints probed answered a handshake (89.4%), on 455 hostsa different slice of the official registry each day, so the share is not compared across days · one attempt per endpoint, an auth challenge counts as an answer · measured from macbook-krmcp-server-liveness v23
How this is made
- Every line comes from one of WITAN's public datasets and links the version its records landed in: provider-incidents, agent-api-observatory, mcp-server-liveness, agent-sdk-releases, agent-tool-releases. They are free to read.
- Incidents are what the providers' own status pages published (Anthropic, OpenAI, GitHub, Cloudflare, npm, Vercel), read every 6 hours; scheduled maintenance is left out. Follow a line's status page link for the provider's text.
- Round-trips and MCP handshakes are WITAN's own measurements from one vantage point, named on each line. A failure there can be the network between us and the service; it is not a verdict on the service. A latency line needs the day's median at least 2× the previous 7 days' and 250 ms more.
- Releases are those of the SDKs and agent tools the datasets follow; pre-releases are left out.
- Days are UTC. A day is rebuilt every hour while it runs and frozen 7 hours after it ends. Model prices are not here: that dataset's collector is off. WITAN is a testnet preview run by one developer.
Earlier days
2026-11-012026-10-312026-10-302026-10-292026-10-282026-10-272026-10-262026-10-252026-10-24