Roadmap

The roadmap prioritizes operator trust, benchmarkable quality, and standalone reliability before any hosted or team-operation story.

Phases 12–13 — Shipped

Agent Harness is now installable and understandable without reading source. Reliability is the core product feature.

Shipped:

• canonical install paths: npm, Docker Compose, source
• harness init, doctor, and demo first-run flows
• examples surface and standalone docs
• compatibility policy and release workflow
• recovery semantics and evidence-rich artifacts
• auditable approval paths and crash coverage

Phase 14 — Current

Make routing and retrieval decisions benchmark-driven.

Focus:

• eval loops
• routing evidence
• semantic-context comparisons
• persisted comparison reports

Phase 15 — Next

Prepare for shared operation without abandoning the standalone product.

Focus:

• storage abstraction
• named operator actions
• export and backup flows
• optional persistent DB backends later