long-running-mcp-tools-must-return-run-id-immediately-not-block
MCP tools wrapping multi-hour research pipelines (like researchclaw_run) must return a run_id immediately and not block the session. Status is polled via a separate researchclaw_status tool; artifacts are retrieved via researchclaw_get_output. Blocking on a long-running pipeline call causes session timeouts and breaks ORACLE Hermes orchestration. This async run_id pattern is the required design for any MCP server wrapping workflows longer than a few seconds.
Related
- mcp-async-pipeline-must-return-run-id-not-block
- oracle
- researchclaw-mcp-async-pattern-run-id-polling
- docker
- mcp-wrapper-pattern-for-long-running-pipelines
- session-scoped-mcp-servers-invisible-to-pre-existing-subagen
- subagent-type-general-purpose-required-for-execution
- permanent-mcp-visibility-vs-session-scoped-for-subagents
- serena-mcp-creates-serena-dir-in-workspace-root
- skill-registry-injection-20k-tokens-per-call