MCP Resources Reference

Cordum MCP resources expose read-only operational context for agents and clients. All resources/read calls return MCP content with:

• contents[0].uri
• contents[0].mimeType (application/json)
• contents[0].text (a JSON-encoded payload string)

Resource Catalog

cordum://jobs/{id}

• Description: Job detail by ID.
• Response format:
  • id, state, topic, tenant
  • submitted_at, completed_at
  • result_ptr, optional result
  • safety_decision, safety_reason, safety_rule_id
  • optional approval metadata
• Example payload (contents[0].text JSON):

{
  "id": "b25c8808-17ad-4114-b013-15552bbf5359",
  "state": "succeeded",
  "topic": "job.demo-guardrails.safe",
  "tenant": "default",
  "submitted_at": "2026-02-13T15:20:10Z",
  "completed_at": "2026-02-13T15:20:11Z",
  "result_ptr": "redis://res:b25c8808-17ad-4114-b013-15552bbf5359",
  "safety_decision": "ALLOW",
  "safety_reason": "matched allow rule",
  "safety_rule_id": "allow-safe-topic"
}

cordum://jobs?status={status}&limit={limit}&cursor={cursor}

• Description: Job list with filters and pagination.
• Query params:
  • status (optional): pending/running/succeeded/failed/cancelled
  • limit (optional): default 20, max 100
  • cursor (optional): pagination cursor
• Response:
  • items: array of jobs
  • next_cursor: cursor for next page (when available)
• Example payload (contents[0].text JSON):

{
  "items": [
    {
      "id": "b25c8808-17ad-4114-b013-15552bbf5359",
      "state": "succeeded",
      "topic": "job.demo-guardrails.safe",
      "submitted_at": "2026-02-13T15:20:10Z"
    },
    {
      "id": "db396e4d-0db6-440a-a0e0-3f2df885c559",
      "state": "running",
      "topic": "job.hello-pack.echo",
      "submitted_at": "2026-02-13T15:19:55Z"
    }
  ],
  "next_cursor": 1739458200000000
}

cordum://workflows/{id}/runs?limit={limit}

• Description: Recent workflow runs for a workflow ID.
• Query params:
  • limit (optional): default 10, max 100
• Response:
  • workflow_id
  • items: array of runs
• Example payload (contents[0].text JSON):

{
  "workflow_id": "demo-guardrails",
  "items": [
    {
      "id": "run-20260213-001",
      "status": "succeeded",
      "started_at": "2026-02-13T15:10:00Z",
      "completed_at": "2026-02-13T15:10:04Z"
    },
    {
      "id": "run-20260213-002",
      "status": "approval_required",
      "started_at": "2026-02-13T15:12:00Z"
    }
  ]
}

cordum://workflows/{id}/runs/{runId}

• Description: Workflow run detail by run ID (scoped by workflow ID).
• Response format:
  • id, workflow_id, status
  • steps array with per-step status and timing
  • optional output, error, and metadata
• Example payload (contents[0].text JSON):

{
  "id": "run-20260213-001",
  "workflow_id": "demo-guardrails",
  "status": "succeeded",
  "steps": [
    {
      "id": "write@1",
      "status": "succeeded",
      "started_at": "2026-02-13T15:10:00Z",
      "completed_at": "2026-02-13T15:10:02Z"
    },
    {
      "id": "safe@1",
      "status": "succeeded",
      "started_at": "2026-02-13T15:10:02Z",
      "completed_at": "2026-02-13T15:10:04Z"
    }
  ]
}

cordum://audit?limit={limit}

• Description: Recent policy audit entries.
• Query params:
  • limit (optional): default 50, max 200
• Response:
  • items: audit entries (timestamp/action/resource/actor metadata)
• Example payload (contents[0].text JSON):

{
  "items": [
    {
      "timestamp": "2026-02-13T15:18:23Z",
      "action": "policy.evaluate",
      "resource": "job:b25c8808-17ad-4114-b013-15552bbf5359",
      "actor": "scheduler",
      "decision": "ALLOW"
    },
    {
      "timestamp": "2026-02-13T15:18:31Z",
      "action": "policy.evaluate",
      "resource": "job:dangerous-001",
      "actor": "scheduler",
      "decision": "DENY"
    }
  ]
}

cordum://health

• Description: System health snapshot.
• Response format:
  • uptime_seconds
  • workers (counts/active)
  • redis, nats, safety_kernel status objects
  • optional build/version metadata
• Example payload (contents[0].text JSON):

{
  "uptime_seconds": 8342,
  "workers": {
    "total": 7,
    "active": 4
  },
  "redis": {
    "status": "ok"
  },
  "nats": {
    "status": "ok"
  },
  "safety_kernel": {
    "status": "ok"
  },
  "version": "dev"
}

cordum://policies

• Description: Active policy bundle summary.
• Response:
  • active_bundles: current bundle summaries
  • current_snapshot_id
  • safety_stance (permissive|balanced|strict)
• Example payload (contents[0].text JSON):

{
  "active_bundles": [
    {
      "id": "[email protected]",
      "name": "demo-guardrails",
      "rule_count": 12,
      "last_updated": "2026-02-13T15:00:00Z"
    }
  ],
  "current_snapshot_id": "snapshot-20260213-150000",
  "safety_stance": "balanced"
}

JSON-RPC Envelope Example

{
  "jsonrpc": "2.0",
  "id": 11,
  "method": "resources/read",
  "params": {
    "uri": "cordum://jobs?status=running&limit=5"
  }
}

{
  "jsonrpc": "2.0",
  "id": 11,
  "result": {
    "contents": [
      {
        "uri": "cordum://jobs?status=running&limit=5",
        "mimeType": "application/json",
        "text": "{\"items\":[...],\"next_cursor\":1739433600000000}"
      }
    ]
  }
}

Pagination Pattern

• List resources return next_cursor when additional pages exist.
• Pass returned next_cursor into the next resources/read URI query.
• Keep limit stable across pages for predictable traversal.

Agent Usage Notes

• Fetch cordum://policies and cordum://health early to build operational context.
• Use cordum://jobs?... before cordum://jobs/{id} to avoid unnecessary deep reads.
• Prefer bounded limits (5-20) for iterative context gathering workflows.