mcp-wrapper-pattern-for-long-running-pipelines
When wrapping long-running pipelines (like AutoResearchClaw’s 23 stages) as MCP servers, the trigger tool (researchclaw_run) should return immediately with a run_id and use polling via a status tool (researchclaw_status). Expose standalone utility tools (verify_citations, verify_claims, search_literature) as independent MCP tools usable outside the full pipeline — this maximizes composability across all agents in the stack.
Related
- oracle
- autoresearchclaw-mcp-server-11-tools-interface-design
- hermes
- autoresearchclaw-mcp-server-integration-pattern
- researchclaw-mcp-server-11-tools-async-run-id-pattern
- long-running-mcp-tools-must-return-run-id-immediately-not-bl
- session-scoped-mcp-servers-invisible-to-pre-existing-subagen
- permanent-mcp-visibility-vs-session-scoped-for-subagents