Deliberated Code Review

When you need a thorough code review from multiple perspectives, delegate to specialist reviewers, cross-vote, and use gemot to find the crux of any disagreements.

When to Use

• PR review where security, performance, and correctness might conflict
• Architecture decisions where subagents return different recommendations
• Any multi-perspective review where you need to know WHY reviewers disagree, not just THAT they disagree

Process

Step 1: Delegate specialist reviews

Use delegate_task to spawn 3 independent reviewers. Each reviews the same code from their expertise:

delegate_task([
  {"goal": "Review this code for security vulnerabilities. Cite specific lines, explain attack vectors, suggest fixes.", "context": "<the code>"},
  {"goal": "Review this code for reliability and correctness. Cite specific lines, explain what breaks, suggest fixes.", "context": "<the code>"},
  {"goal": "Review this code for performance issues. Cite specific lines, estimate impact, suggest fixes.", "context": "<the code>"}
])

Step 2: Create a gemot deliberation

Call mcp_gemot_deliberation with:

• action: "create"
• topic: a one-line summary of what's being reviewed (e.g., "Payment processing code review")
• template: "review"
• group_id: a consistent identifier for the project or workflow (e.g., "myproject-reviews")

Step 3: Submit each reviewer's findings

For each subagent result, call mcp_gemot_participate with:

• action: "submit_position"
• deliberation_id: from step 2
• agent_id: the reviewer role (e.g., "security-reviewer")
• content: the reviewer's full findings

Step 4: Cross-vote

Call mcp_gemot_participate with action: "get_positions" to fetch all positions. Then have each reviewer vote on the other reviewers' findings:

For each position, each OTHER reviewer calls mcp_gemot_participate with:

• action: "vote"
• deliberation_id: from step 2
• agent_id: the voting reviewer's ID
• position_id: the position being voted on
• value: 1 (agree with this finding), 0 (no opinion), or -1 (disagree)

This feeds gemot's clustering and crux detection with real signal about where reviewers actually agree vs disagree.

Step 5: Analyze

Call mcp_gemot_analyze with action: "run". Wait for completion by polling mcp_gemot_deliberation with action: "get" until status returns to "open" (poll every 5 seconds, allow up to 5 minutes).

Step 6: Read the results

Call mcp_gemot_analyze with action: "get_result" for the full analysis and mcp_gemot_participate with action: "get_context" and any reviewer's agent_id. Report to the user:

1. What all reviewers agree on — consensus statements (things to definitely fix)
2. Where they disagree — cruxes (the specific claims that divide them), with controversy scores
3. Proposed compromise — gemot's suggested resolution
4. What the user should decide — value judgments that can't be resolved with evidence

Step 7 (optional): Commit to fixes

If the user agrees on what to fix, call mcp_gemot_decide with action: "commit" for each agreed fix:

• deliberation_id: from step 2
• agent_id: "team" or the user's identifier
• statement: the specific fix commitment (e.g., "Add SELECT FOR UPDATE to prevent race condition")

These commitments are tracked and can be verified in future reviews via mcp_gemot_decide with action: "reputation".

Step 8 (optional): Refine

If the user wants deeper analysis, have each reviewer call mcp_gemot_participate with action: "get_context" to see the cruxes, then submit a refined position addressing the disagreements. Re-analyze. Cruxes get more specific each round.

Example output to user

All reviewers agree: The balance check on line 4 has a TOCTOU race condition. The notify_payment_service call after commit has no error handling. The amount parameter should be Decimal, not float.

>

They disagree on: Whether to use SELECT FOR UPDATE (security wants it for atomicity, performance says it creates lock contention at scale). Controversy: 67%.

>

Gemot's take: Use SELECT FOR UPDATE with a short lock timeout. The security risk of the race condition outweighs the performance cost at your current scale.

>

You decide: Should notify_payment_service be inside or outside the transaction? Security says inside (atomicity), reliability says outside (don't hold DB connections during external calls). This is a value crux — depends on your failure mode preferences.

Gemot tools (6 grouped tools)

All gemot MCP tools use the mcp_gemot_ prefix with an action parameter:

• mcp_gemot_deliberation — actions: create, get, list, list_by_group, list_by_agent, delete, set_template, export
• mcp_gemot_participate — actions: submit_position, publish_position, vote, get_positions, get_context, withdraw
• mcp_gemot_analyze — actions: run, get_result, cancel, propose_compromise, reframe, challenge, dispute_crux
• mcp_gemot_decide — actions: commit, get_commitments, fulfill, break, reputation
• mcp_gemot_coordinate — actions: delegate, invite, generate_join_code, join
• mcp_gemot_admin — actions: report_abuse, get_audit_log, list_templates, get_votes