DEDUPED Hot AI repos. Real use cases. What people build. Get the Monday 5

How to get a code review from three models that must agree with Tstack

Use the council skill to have models from three providers review a change until they reach agreement.

Short answer: Install Tstack as above; then set which models sit on your council in your CLAUDE.md or AGENTS.md; make sure your harness can run subagents on other providers' models. All steps.

Who it is forA senior engineer who does not trust a single model's review on risky changes.

The problemOne model's review has blind spots, and you cannot tell which findings are real.

The outcomeA review where three models from different providers loop until they agree, so the remaining findings are more trustworthy.

What you need first

  • Claude Code or Codex
  • An authenticated GitHub CLI (gh)
  • python3 for babysit-pr's pr-wait script
  • council and sol-loop need subagents on other providers' models (T3 Code or a harness that can run them)

Steps

  1. Install Tstack as above
  2. Set which models sit on your council in your CLAUDE.md or AGENTS.md
  3. Make sure your harness can run subagents on other providers' models
  4. Say council on the change you want reviewed
  5. Act on the findings the models agree on

Prompt for your agent

Use the council skill from t3dotgg/tstack to review the diff on this branch with models from three providers, loop until they agree, and list only the agreed findings with file and line.

Caveat. council and sol-loop need subagents on other providers' models; without T3 Code you need a harness that can run them.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for t3dotgg/tstack and "Get a code review from three models that must agree"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer