Config file
optamus.toml in the repo pins model, parallelism, and tool allowlists for the whole team.
One command starts a run. Flags, exit codes, and JSON output make it scriptable from the first install.
$ optamus run "migrate the billing tests to vitest"
▸ Reading workspace · 42 files
▸ Running suite · 38 passed, 2 flaky
✓ Patched src/billing/*.test.ts (+214 −188)
✓ All green — run/1a2b finished in 2m 04s Everything the CLI shows you is a flag away in machine form. Pipe it, branch on it, or drop it into a Makefile.
$ optamus run "triage open issues" --limit 20 --json | jq '.state'
"completed" Install the CLI in a job, point it at a checkout, and let it open a pull request. The same agent you run locally — same flags, same output.
- name: Run agent
run: |
curl -fsSL https://optamus.cloud/install.sh | sh
optamus run "fix the failing e2e tests" \
--patch out.patch --receipt run.json
- name: Open pull request
uses: peter-evans/create-pull-request@v6
with: { patch: out.patch } optamus.toml in the repo pins model, parallelism, and tool allowlists for the whole team.
Fan one command into four runs, each in its own worktree, since 1.4.
Settled, interrupted, failed — three codes, easy to branch on.
Start in the shell, watch it in the app. The run is the same run.
Install the desktop app or the CLI, run one command, and review the first diff.