# Agent Fixboard > Step-by-step fix guides for real AI agent failures. Every guide links to its public source and carries a fix status (Fixed, Workaround, Open or Unverified), re-checked against the upstream GitHub issue. 25 guides. Last status check: 2026-10-15. Fixed = fixed upstream by a merged PR or commit (evidence linked). Open = the upstream issue is still open. Workaround = upstream closed as not planned, stale or duplicate; a workaround is given. Unverified = no proof yet. ## Feeds - [RSS: new and re-checked fixes](https://agentfixboard.com/feed.xml) - [Sitemap](https://agentfixboard.com/sitemap.xml) ## Stacks - LangGraph: 5 guides - browser-use: 4 guides - AutoGen: 3 guides - MCP: 3 guides - LangChain: 2 guides - MCP Python SDK: 2 guides - n8n: 2 guides - Aider: 1 guide - Continue: 1 guide - CrewAI: 1 guide - Hermes Agent: 1 guide ## Status counts - Unverified: 9 - Workaround: 8 - Fixed: 4 - Open: 4 ## Guides - [Coding agent stuck on 'did not conform to the edit format'](https://agentfixboard.com/guides/aider-edit-format-loop): Aider. Status: Unverified, checked 2026-10-15. - ['str' object has no attribute 'get' when the chat finishes](https://agentfixboard.com/guides/autogen-str-has-no-attribute-get): AutoGen. Status: Fixed, checked 2026-10-15. - [Adding a second agent to the quick-start makes it loop forever](https://agentfixboard.com/guides/autogen-core-cascading-messages-loop): AutoGen. Status: Unverified, checked 2026-10-15. - [Graph flow errors with 'unable to find next speaker' when a loop has tools](https://agentfixboard.com/guides/autogen-graphflow-loop-needs-exit): AutoGen. Status: Unverified, checked 2026-10-15. - [Browser agent can't pick an option from a dropdown](https://agentfixboard.com/guides/browser-use-cannot-select-dropdown): browser-use. Status: Fixed, checked 2026-10-15. - [Browser agent keeps emitting empty actions until max steps](https://agentfixboard.com/guides/browser-use-empty-action-loop): browser-use. Status: Unverified, checked 2026-10-15. - [Browser agent keeps going after it already called done](https://agentfixboard.com/guides/browser-use-does-not-stop-after-done): browser-use. Status: Unverified, checked 2026-10-15. - [Click on a missing element is reported as success](https://agentfixboard.com/guides/browser-use-missing-element-reported-as-success): browser-use. Status: Open, checked 2026-10-15. - [Tool calls fail because arguments come back as XML-like text](https://agentfixboard.com/guides/tool-args-in-xml-instead-of-json): Continue. Status: Workaround, checked 2026-10-15. - [Every tool call arrives with empty arguments and fails validation](https://agentfixboard.com/guides/crewai-tool-arguments-dropped-empty-dict): CrewAI. Status: Workaround, checked 2026-10-15. - [Error messages name the symptom, not the cause](https://agentfixboard.com/guides/hermes-agent-misleading-error-messages): Hermes Agent. Status: Workaround, checked 2026-10-15. - [JSON output parser returns None instead of raising on bad JSON](https://agentfixboard.com/guides/langchain-json-parser-returns-none): LangChain. Status: Workaround, checked 2026-10-15. - [Parser fails when the model wraps JSON in ```JSON instead of ```json](https://agentfixboard.com/guides/langchain-uppercase-json-fence): LangChain. Status: Workaround, checked 2026-10-15. - ["tool_use ids were found without tool_result blocks" on a saved thread](https://agentfixboard.com/guides/langgraph-tool-use-without-tool-result-checkpoint): LangGraph. Status: Unverified, checked 2026-10-15. - [Agent repeats the same tool call until it hits the recursion limit](https://agentfixboard.com/guides/langgraph-agent-loops-until-recursion-limit): LangGraph. Status: Workaround, checked 2026-10-15. - [Cancelled async tool leaves history invalid even with handle_tool_errors=True](https://agentfixboard.com/guides/langgraph-cancelled-tool-breaks-history): LangGraph. Status: Open, checked 2026-10-15. - [Setting recursion_limit to 10000 silently does nothing](https://agentfixboard.com/guides/langgraph-recursion-limit-10000-ignored): LangGraph. Status: Workaround, checked 2026-10-15. - [Token limit exceeded because retriever output piles up in chat history](https://agentfixboard.com/guides/langgraph-context-overflow-from-tool-messages): LangGraph. Status: Unverified, checked 2026-10-15. - [Fetch tool can be steered to localhost or cloud metadata](https://agentfixboard.com/guides/mcp-fetch-can-reach-internal-network): MCP. Status: Fixed, checked 2026-10-15. - [Filesystem server says 'path outside allowed directories' on the first call](https://agentfixboard.com/guides/mcp-filesystem-roots-race): MCP. Status: Fixed, checked 2026-10-15. - [npx-based server shows zero tools when Node comes from nvm](https://agentfixboard.com/guides/mcp-server-fails-with-nvm): MCP. Status: Unverified, checked 2026-10-15. - [Stateless MCP server logs ERROR every time a client disconnects](https://agentfixboard.com/guides/mcp-stateless-server-error-logs-on-disconnect): MCP Python SDK. Status: Workaround, checked 2026-10-15. - [Streamable HTTP client keeps reconnecting past its retry limit](https://agentfixboard.com/guides/mcp-client-reconnects-past-retry-budget): MCP Python SDK. Status: Open, checked 2026-10-15. - [AI Agent tools fail with 'schema must be type object, got None' after upgrading](https://agentfixboard.com/guides/n8n-tool-schema-none-after-upgrade): n8n. Status: Unverified, checked 2026-10-15. - [One bad tool call ends the whole agent run](https://agentfixboard.com/guides/n8n-schema-invalid-tool-call-kills-run): n8n. Status: Open, checked 2026-10-15. ## Optional - [Sponsor](https://agentfixboard.com/sponsor) - [Terms](https://agentfixboard.com/terms) - [Content removal](https://agentfixboard.com/takedown)