Agent failure knowledge base

Learn from the failures of other AI agents.

Known agent failures. Known causes. Repairs that were actually tested — and the ones that did not work.

FAILURE→ SYMPTOMS→ CONTEXT→ ROOT CAUSE→ REMEDIATION ATTEMPTS→ VERIFICATION→ OUTCOME→ RECURRENCE

Search by symptoms Use the API

What is in the database

115

published cases (deduplicated incidents)

229

linked public sources

183

documented remediation attempts

documented failed repairs

Every number above is counted from the published records (ds-c1e9fc067ec7). Nothing is estimated. Breakdown by category →

Evidence first

Verbatim, verified quotes

Every claim is tied to a verbatim quote from a public source and labelled VERIFIED_FACT, SUPPORTED_INTERPRETATION, REPORTED_CLAIM, INFERENCE or UNKNOWN.

Statuses are earned

A fix is VERIFIED SUCCESS only with independent confirmation that the failure is gone. A merged fix with a regression test is TESTED. A proposal is SUGGESTED. A cause without project confirmation stays LIKELY or UNKNOWN.

Similarity, not diagnosis

Search returns documented cases with similar symptoms and explains every match. It never claims your agent has the same defect. A search with no match is free.

Global by design

Query in any language. Original error messages, quotes and identifiers are preserved; a Russian or Japanese query can find an English or Chinese case. Language is never treated as geography.

Well-documented cases

AE-3FRMS134 · LangGraphAfter upgrading LangGraph, interrupt() inside an agent tool is not surfaced when streaming with stream_mode="values"INTEGRATION FAILUREcause: VERIFIEDoutcome: RESOLVED UNVERIFIEDconfidence: HIGH AE-49973DV6 · LangChainAgent's invalid-tool-call repair leaves an orphaned tool_result, so Anthropic-backed threads fail with a permanent 400TOOL FAILUREcause: VERIFIEDoutcome: RESOLVED UNVERIFIEDconfidence: HIGH AE-52RYEZYX · OpenAI Agents SDK (Python)Concurrent first use of an agent conversation session creates two conversations and splits the session historyCONCURRENCY FAILUREcause: VERIFIEDoutcome: RESOLVED UNVERIFIEDconfidence: HIGH AE-GDZEYNCQ · LlamaIndexRestricted safe_eval of LLM-generated code is bypassed via prompt injection, executing arbitrary code (bypass of an earlier fix)SECURITY FAILUREcause: LIKELYoutcome: RESOLVED UNVERIFIEDconfidence: MEDIUM AE-GWWP1DFD · Model Context Protocol reference servers (memory)Memory MCP server corrupts or loses its knowledge-graph file when an agent issues several memory tool calls in one turnMEMORY FAILUREcause: VERIFIEDoutcome: RESOLVED UNVERIFIEDconfidence: HIGH AE-H8AK9466 · Semantic Kernel (Python)Agent wrapper drops the source filename (title) from url_citation annotations, leaving citations traceable only to opaque document idsEVIDENCE FAILUREcause: VERIFIEDoutcome: RESOLVED UNVERIFIEDconfidence: HIGH

For agents

curl -s -X POST https://agenterrors.online/api/v1/search \
  -H 'content-type: application/json' \
  -d '{"query":"My agent retries the same HTTP request forever after a timeout"}'
# no match → 200 {"no_match":true} (free)
# match    → 402 with a free preview; pay 0.018 USDC via x402, retry with PAYMENT-SIGNATURE

Also known as: AI agent errors · AI agent failures · agent failure database · agent incident database · agent bug knowledge base · agent remediation evidence · AI agent troubleshooting · agent failure patterns · agent root causes · agent repair outcomes.