ResearchClaw v0.4.0 — Defer to Dedicated Design Session
Decision
Defer v0.4.0 upgrade. Preserve backup at /opt/enterprise-backup/researchclaw/pre-upgrade-20260415 (92M). Schedule a separate design session with decision_matrix for HITL tool contract design.
Rationale
Upstream mcp_server is a stdio 6-tool placeholder scaffold. Our production is a streamable-HTTP 11-tool wrapper + 6 new HITL tools = 17+ tools that must be surfaced as MCP after rebuild. Config schema grew 6.3×. Naive file-copy would regress Law #10 (Full Capability). This is no longer a routine upgrade — it’s a re-architecture with tool-contract design decisions (which HITL tools belong in MCP vs internal API, how to surface v0.4.0 new capabilities, what breaks our Claude Code integration). Needs structured ST decision_matrix + first_principles on tool ontology. Current v0.3.1 is production-stable — no urgency.
Alternatives Rejected
Naive file-copy upgrade (rejected): Would lose 11-tool HTTP wrapper + 6 HITL tools. Regresses Law #10. Full rewrite in this session (rejected): No design time for tool-contract decision, too risky without decision_matrix. Defer with backup preserved (chosen): Zero risk, production stays on v0.3.1, clean architectural rebuild planned.
Outcome
Pending
Related
- docker
- oracle
- hermes-v090-port-individual-features-option-c-instead-of-ful
- hermes
- 2026-04-04-oracle-001-self-architecture-analysis
- researchclaw-v040-streamable-http-wrapper-incompatible
- researchclaw-v040-http-wrapper-has-no-upstream-counterpart
- researchclaw-v040-schema-grew-6x-upstream-scaffold-incompati