Daily digest · 2026-10-27 (UTC)

What broke for AI agents on 27 Oct 2026

14 incidents on 4 status pages.

Final. Built 2026-10-31 11:03 UTC, 7 hours after the day ended; this page will not change.

What broke

1. Anthropic: “Elevated errors for Claude Sonnet 5.5”major impact · opened 09:14 UTC · resolved 09:56 UTC (42 minutes) · Claude API (api.anthropic.com) | Claude Code | Claude Cowork | claude.aistatus page ↗provider-incidents v68
2. Anthropic: “Elevated errors for Claude Sonnet 5.5”major impact · opened 15:09 UTC · resolved 15:50 UTC (41 minutes) · Claude API (api.anthropic.com) | Claude Code | Claude Cowork | claude.aistatus page ↗provider-incidents v69
3. Cloudflare: “ERR_TIMED_OUT and ERR_EMPTY_RESPONSE being observed by Cloudflare Customers”major impact · opened 12:28 UTC · resolved 15:12 UTC (2 h 44 min)status page ↗provider-incidents v69
4. Cloudflare: “Cache Purging Delays Affecting Requests Routed Through EEUR”minor impact · opened 15:10 UTC · resolved 16:14 UTC (64 minutes) · CDN Cache Purge | Stockholm, Sweden - (ARN)status page ↗provider-incidents v69
5. Cloudflare: “Cloudflare Stream performance issues impacting certain Live Broadcasts and On-Demand Uploads”minor impact · opened 16:10 UTC · resolved 17:02 UTC (51 minutes) · Streamstatus page ↗provider-incidents v69
6. Cloudflare: “Delayed, missed, and duplicated Workers Cron Triggers”minor impact · opened Oct 26 21:48 UTC · resolved 01:15 UTC (3 h 26 min) · Workersstatus page ↗provider-incidents v66
7. Cloudflare: “Some customers may notice their Pages and/or Workers builds are intermittently failing.”minor impact · opened 08:04 UTC · resolved 08:17 UTC (13 minutes) · Pages | Workers Buildsstatus page ↗provider-incidents v70
8. Cloudflare: “We are seeing some customers having issues with Workers Builds”minor impact · opened 11:38 UTC · resolved 12:55 UTC (77 minutes) · Workers Buildsstatus page ↗provider-incidents v68
9. Cloudflare: “Elevated Cloudflare Access API Errors”no impact level set · opened 21:13 UTC · resolved 21:58 UTC (44 minutes)status page ↗provider-incidents v70
10. Cloudflare: “Issues with AI Security for Apps”no impact level set · opened 16:31 UTC · resolved 17:09 UTC (39 minutes) · Workers AIstatus page ↗provider-incidents v69
11. Cloudflare: “Network Error Log Processing Unavailable”no impact level set · opened 22:21 UTC · resolved 22:58 UTC (37 minutes)status page ↗provider-incidents v70
12. GitHub: “Elevated 404s for GitHub App s2s tokens in git/API”no impact level set · opened 08:00 UTC · resolved 08:00 UTC (0 minutes)status page ↗provider-incidents v70
13. OpenAI: “6.1 Sol Degraded Model Performance”the feed gives no impact level · the feed gives no start time · resolved 16:20 UTC · Batch | Chat Completions | Responsesstatus page ↗provider-incidents v69
14. OpenAI: “Elevated error rates for most hosted Code Reviews and Security Reviews on private GitHub repositories”the feed gives no impact level · first seen by our reader 08:14 UTC (the feed gives no start time) · resolved 08:30 UTCstatus page ↗provider-incidents v68

Shipped

No new major version of the SDKs and tools we follow.

23 other releases on the day

1. @anthropic-ai/claude-agent-sdk 0.3.312 (npm)released 18:37 UTC · previous 0.3.311agent-sdk-releases v84
2. ai 7.0.155 (npm)released 19:37 UTC · previous 7.0.154agent-sdk-releases v84
3. anthropics/claude-code v2.1.312published 20:24 UTC · previous v2.1.311release notes ↗agent-tool-releases v85
4. BerriAI/litellm v1.103.10published 19:55 UTC · previous v1.106.4release notes ↗agent-tool-releases v85
5. BerriAI/litellm v1.104.8published 19:57 UTC · previous v1.103.10release notes ↗agent-tool-releases v85
6. BerriAI/litellm v1.105.6published 19:58 UTC · previous v1.104.8release notes ↗agent-tool-releases v85
7. BerriAI/litellm v1.106.5published 19:58 UTC · previous v1.105.6release notes ↗agent-tool-releases v85
8. crewAIInc/crewAI 1.15.31published 21:36 UTC · previous 1.15.30release notes ↗agent-tool-releases v86
9. crewAIInc/crewAI 1.15.32published 23:04 UTC · previous 1.15.31release notes ↗agent-tool-releases v86
10. google-gemini/gemini-cli v0.66.0published 20:26 UTC · previous v0.65.0release notes ↗agent-tool-releases v85
11. langchain-ai/langchain langchain-anthropic==1.7.7published 23:48 UTC · previous langchain-anthropic==1.7.6release notes ↗agent-tool-releases v86
12. litellm 1.106.5 (pypi)released 19:44 UTC · previous 1.106.4agent-sdk-releases v84
13. ollama/ollama v0.40.12published 00:07 UTC · previous v0.40.11release notes ↗agent-tool-releases v86
14. openai 3.40.0 (pypi)released 15:38 UTC · previous 3.39.0agent-sdk-releases v84
15. openai 7.45.1 (npm)released 15:49 UTC · previous 7.45.0agent-sdk-releases v84
16. openai/openai-node v7.45.1published 15:37 UTC · previous v7.45.0release notes ↗agent-tool-releases v85
17. openai/openai-python v3.40.0published 15:37 UTC · previous v3.39.0release notes ↗agent-tool-releases v85
18. vercel/ai @ai-sdk/[email protected]published 19:37 UTC · previous @ai-sdk/[email protected]release notes ↗agent-tool-releases v85
19. vercel/ai @ai-sdk/[email protected]published 23:12 UTC · previous @ai-sdk/[email protected]release notes ↗agent-tool-releases v86
20. vercel/ai @ai-sdk/[email protected]published 19:37 UTC · previous @ai-sdk/[email protected]release notes ↗agent-tool-releases v85
21. vercel/ai @ai-sdk/[email protected]published 19:37 UTC · previous @ai-sdk/[email protected]release notes ↗agent-tool-releases v85
22. vercel/ai [email protected]published 19:28 UTC · previous [email protected]release notes ↗agent-tool-releases v85
23. vercel/ai [email protected]published 19:33 UTC · previous [email protected]release notes ↗agent-tool-releases v85

Measured on the day

1. 184 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 v112
2. 465 of 500 remote MCP endpoints probed answered a handshake (93.0%), on 461 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 v26

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