Frontier Model Tools

Rigour v2.14+ includes specialized MCP tools designed for next-generation AI models like Claude Opus 4.6 and GPT-5.3-Codex that support:

• Multi-agent collaboration - Multiple AI agents working on the same codebase
• Long-running sessions - Extended coding sessions (hours/days)
• Quality checkpointing - Periodic quality assessments with drift detection

Tool Overview

Tool	Purpose	Key Feature
rigour_agent_register	Register agent in session	Scope conflict detection
rigour_agent_deregister	Remove agent from session	Clean session management
rigour_checkpoint	Record quality checkpoint	Drift detection
rigour_handoff	Initiate task handoff	Multi-agent coordination
rigour_handoff_accept	Accept pending handoff	Verified handoff flow

---

rigour_agent_register

Register an agent at the START of execution to claim task scope and enable cross-agent conflict detection.

Parameters

Parameter	Type	Required	Description
cwd	string	✅	Absolute path to project root
agentId	string	✅	Unique identifier (e.g., agent-frontend)
taskScope	string[]	✅	Glob patterns for claimed files

Example

{
  "cwd": "/path/to/project",
  "agentId": "opus-api-agent",
  "taskScope": ["src/api/**", "tests/api/**"]
}

Response

✅ AGENT REGISTERED: "opus-api-agent" claimed scope: src/api/**, tests/api/**

Active agents in session: 1

Conflict Detection

If another agent already claims overlapping scope:

⚠️ SCOPE CONFLICTS DETECTED:
  - opus-frontend-agent also claims "src/shared/**"

Consider coordinating with other agents or narrowing your scope.

---

rigour_agent_deregister

Remove an agent from the session when work is complete or scope needs to be released.

Parameters

Parameter	Type	Required	Description
cwd	string	✅	Absolute path to project root
agentId	string	✅	ID of agent to deregister

Example

{
  "cwd": "/path/to/project",
  "agentId": "opus-api-agent"
}

---

rigour_checkpoint

Record a quality checkpoint during long-running sessions. Call every 15-30 minutes to enable drift detection and quality monitoring.

Parameters

Parameter	Type	Required	Description
cwd	string	✅	Absolute path to project root
progressPct	number	✅	Estimated progress (0-100)
summary	string	✅	Brief description of recent work
qualityScore	number	✅	Self-assessed quality (0-100)
filesChanged	string[]	❌	List of modified files

Example

{
  "cwd": "/path/to/project",
  "progressPct": 45,
  "summary": "Implemented user authentication endpoints",
  "qualityScore": 88,
  "filesChanged": ["src/api/auth.ts", "tests/auth.test.ts"]
}

Quality Warnings

The checkpoint system provides warnings when:

1. Quality Below Threshold (< 80%):

   ⚠️ Quality score 65% is below threshold 80%
   ⛔ QUALITY BELOW THRESHOLD: Consider pausing and reviewing recent work.

2. Drift Detected (quality degrading over time):

   ⚠️ Drift detected: quality dropped from avg 88% to 70%

---

rigour_handoff

Initiate a task handoff to another agent. Use when delegating a subtask or completing your scope.

Parameters

Parameter	Type	Required	Description
cwd	string	✅	Absolute path to project root
fromAgentId	string	✅	ID of initiating agent
toAgentId	string	✅	ID of receiving agent
taskDescription	string	✅	Description of the task
filesInScope	string[]	❌	Relevant files
context	string	❌	Additional context

Example

{
  "cwd": "/path/to/project",
  "fromAgentId": "opus-api-agent",
  "toAgentId": "opus-frontend-agent",
  "taskDescription": "Integrate the new auth endpoints with the login form",
  "filesInScope": ["src/api/auth.ts", "src/components/LoginForm.tsx"],
  "context": "Auth endpoints return JWT tokens in the response body"
}

---

rigour_handoff_accept

Accept a pending handoff. This verifies you are the intended recipient.

Parameters

Parameter	Type	Required	Description
cwd	string	✅	Absolute path to project root
handoffId	string	✅	ID of the handoff to accept
agentId	string	✅	Your agent ID (must match toAgentId)

Example

{
  "cwd": "/path/to/project",
  "handoffId": "handoff-1707234567890",
  "agentId": "opus-frontend-agent"
}

---

Multi-Agent Workflow Example

sequenceDiagram
    participant A as Agent A (API)
    participant B as Agent B (Frontend)
    participant R as Rigour MCP

    A->>R: rigour_agent_register(agentId: "agent-a", scope: ["src/api/**"])
    R-->>A: ✅ Registered
    
    Note over A: Works on API endpoints...
    
    A->>R: rigour_checkpoint(progress: 50, quality: 90)
    R-->>A: 📍 Checkpoint recorded
    
    A->>R: rigour_handoff(to: "agent-b", task: "Integrate endpoints")
    R-->>A: 🤝 Handoff initiated
    
    A->>R: rigour_agent_deregister("agent-a")
    R-->>A: ✅ Deregistered
    
    B->>R: rigour_handoff_accept(handoffId, "agent-b")
    R-->>B: ✅ Accepted
    
    B->>R: rigour_agent_register("agent-b", scope: ["src/components/**"])
    R-->>B: ✅ Registered

---

Best Practices

1. Always register first - Call rigour_agent_register at session start
2. Checkpoint regularly - Every 15-30 minutes for long sessions
3. Be honest with quality scores - Artificially high scores trigger drift detection
4. Narrow your scope - Claim only files you'll actually modify
5. Complete the handoff flow - Use handoff_accept before claiming scope
6. Deregister when done - Release scope for other agents