What's next

The Meshfleet roadmap. Shipped versions, the next release, and the path to v1.0. Dates are aspirational; everything is best-effort.

shipped

v0.4.0

— Resilience
  • ✓ Per-fleet timeouts (overrides global default)
  • ✓ Structured event log (NDJSON)
  • ✓ list_fleets MCP tool
shipped

v0.5.0

— Observability
  • ✓ CLI inspector (npx agent-mesh inspect)
  • ✓ getFleetMetrics with 9 metric fields
  • ✓ Capability-based routing improvements
shipped

v0.5.1

— Health + read-side rate limiting
  • ✓ ping MCP tool (liveness check)
  • ✓ get_health MCP tool (full report)
  • ✓ Read-side rate limiting (600 reads/hour per IP)
next

v0.6.0

— Real-time + fleet templates
  • ◐ Heartbeat watchdog (auto-fail agents that miss N heartbeats)
  • ◐ Fleet templates (save / list / spawn_from_template)
  • ◐ Automatic retry with exponential backoff
planned

v0.7.0

— Smart routing
  • ○ Embedding-based capability matching (local MiniLM model)
  • ○ Skill taxonomy (frontend > react > nextjs hierarchy)
  • ○ Multi-agent route_work (return top N matches for fan-out)
  • ○ Routing feedback loop (learn from failures)
planned

v1.0.0

— Stable
  • ○ Schema versioning in the ledger
  • ○ Backward compatibility matrix (1.x reads v0.9.x fleets)
  • ○ Performance benchmarks (sub-100ms overhead per agent spawn)
  • ○ Distributed ledger option (libSQL behind a feature flag)
  • ○ npm publish (@meshfleet/agent-mesh)
  • ○ Optional auth token for MCP

Have an idea?

If something's missing or you'd like something prioritized, open a GitHub issue with the enhancement label. We especially want to hear about:

Open an issue