MCP (Model Context Protocol)
The MCP server provides programmatic access to Zen Mesh data through the Model Context Protocol.
Current Status
- Server implemented at
src/saas/mcp/ in zen-platform
- Deployment: K8s deployment exists but disabled in sandbox (
mcp.enabled: false)
- Not production-live
- No evidence/proof/readiness tools yet — existing tools cover API keys, deliveries, webhooks
| Tool | Description | Read-Only |
|---|
list_api_keys | List API keys | ✅ |
create_api_key | Create API key | ❌ |
revoke_api_key | Revoke API key | ❌ |
list_deliveries | List deliveries | ✅ |
get_delivery | Get delivery details | ✅ |
list_webhooks | List webhooks | ✅ |
get_webhook | Get webhook details | ✅ |
get_delivery_stats | Get delivery statistics | ✅ |
| Tool | Description | Status |
|---|
get_runtime_status | Runtime convergence overview | planned |
get_trust_proof | Trust proof by ID | planned |
get_compliance_mapping | Compliance framework mappings | planned |
explain_non_claims | Non-claims by category | planned |
summarize_readiness | Readiness summary | planned |
Intended Resources
| URI | Description | Status |
|---|
zen://capabilities | All capabilities with proof status | planned |
zen://runtime/proofs | Runtime proof ledger | planned |
zen://trust/proofs | Trust proof ledger | planned |
zen://compliance/map | Compliance feature graph | planned |
zen://non-claims | Non-claims by category | planned |
zen://readiness | Readiness and blocker summary | planned |
Design Principles
- Read-only for evidence tools: Proposed tools are read-only
- Rate-limited: 50 req/s, burst 100
- Database-backed: Uses persistent storage
- Tenant-scoped: API key authentication required
Non-Claims
- Not production-live — disabled in sandbox
- Evidence tools are planned — not yet implemented
- Existing mutating tools (create/revoke API key) preserved but out of evidence scope
- No compliance certification via MCP