oh-my-t3

Give the agent a goal.
Get back reviewed work.

oh-my-t3 adds plans, roles, task briefs, and reviews to the agents that you run in T3 Code. The agent plans and builds the change in its own thread. Then two T3 child tasks check it: a model of the other family reviews it, and a QA task runs the checks.

npx oh-my-t3
Other ways to install
omt-ulwOne goal, one change

Example: ulw add OAuth to the API

The agent plans and builds three units in its own thread. A Codex child reviews Claude's change, and a QA child runs the checks. A finding goes back for a fix, and the second pass shows in yellow.

omt-swarmOne plan, many tracks

Example: omt-swarm build the API plan

Three tracks from one plan. Each builds and checks in its own T3 worktree thread, at its own speed. Then one review, one merge, one QA run, and one pull request.

Install

  1. 1 Get the omt command

    bun add -g oh-my-t3

    Also needs Node 22 or later.

  2. 2 Install the skills

    omt install --hooks
omt doctor

    For every project. A project then needs no install of its own.

Then restart your T3 agent sessions. --hooks adds the ulw keyword to Claude Code. The install guide lists each file and the uninstall.

How it works

  1. 1 Give a goal

    Type $omt-ulw and the goal in a T3 thread.

  2. 2 Plan

    The agent asks its questions, then writes a plan in .omt/plans/.

  3. 3 Build

    The agent builds the change unit by unit in its own thread.

  4. 4 Review

    A child task on the other model family reviews the change.

  5. 5 Test and report

    A QA child task runs every check, and the agent reports with evidence.

  6. Done

The states of one change

  • Planned, then built, then in review.
  • In review, approve: the change is tested, then done.
  • In review, request changes: the change goes back to built for a fix, and is reviewed again.
  • In review, block: the change is blocked until the plan changes.
A review can send the change back for a fix, or block it until the plan changes.

Two parts

The skills tell the agent how to plan, build, review, and test.

The omt CLI does the fixed steps: the config, the model for each role, the task briefs, and the ledger.

No runtime dependencies or telemetry. Update checks are opt-in.

You can read the plan, the task count, and each review verdict in files under .omt/.

.omt/plans/
The plan of each run, split into units.
.omt/ledger.jsonl
One line for each task start and each result.
.omt/config.jsonc
The profile: the models, the checks, and the limits.

A plan and its ledger stay in .omt/ after context compaction and across sessions, so one plan can run for days.

Read the walkthrough

From the walkthrough of two real runs on the omt repository, one station per unit:

One run had 4 units and used 29 child tasks in 2 hours 20 minutes.

The other had 10 units and used 128 child tasks over 5 days.

Skills

The agent asks its questions, plans, and builds the change in its own thread. Then one review child fixes small findings and reports the rest, and one QA child runs the checks.

Who it is for

Use omt if

You use T3 Code and want an agent to plan and finish work that has several steps. With both Claude Code and Codex, Claude and Codex review each other's work.

Do not use omt if

  • You do not use T3 Code. The workflows need a T3 Code thread to start child tasks.
  • You want a quick edit with no plan and no review. omt plans, reviews, and tests even a small task, with at least 2 child tasks.
  • You want a hosted service or a scheduler that runs in the background. omt has neither.

Requirements

  • The latest T3 Code release.
  • Claude Code, Codex, or both, enabled in T3. Design and UI tasks can run with either provider.
  • git.
  • Node 22 or later, except for the release binary.

This is version 0.5.0. The config format and the CLI output can change.

Credits

oh-my-t3 uses ideas from oh-my-openagent (fallback chains, the ultrawork loop, plan reviews), oh-my-pi (model roles and the review recipe), and T3 Code (orchestrator v2, which runs the child tasks). It copies no code or prompts from oh-my-openagent.

Run omt once, with no install.

npx oh-my-t3