locus iterate

Re-execute tasks based on PR review feedback. Closes the feedback loop between code review and implementation.

Re-execute AI agents on pull requests that have received review feedback. This closes the feedback loop: run produces a PR, review (or a human) leaves comments, and iterate addresses them.

Usage

locus iterate [options]
locus iterate <issue-number>
locus iterate --pr <pr-number>
locus iterate --sprint

---

Options

Flag	Description
--pr <number>	Iterate on a specific PR by number
--sprint	Iterate on all open agent PRs that have feedback
--dry-run	Show what would be iterated without executing
--model <name>	Override the AI model

---

Modes

All Agent PRs (default)

When called without arguments, finds all open PRs labeled agent:managed that have non-bot comments, and iterates on each one.

locus iterate

Specific PR

Iterate on a single PR by number.

locus iterate --pr 15

The command fetches all comments on the PR and passes them as context to the AI agent, which then makes the requested changes and pushes an update.

By Issue Number

Pass an issue number to automatically find its associated PR and iterate.

locus iterate 42

Locus searches for open PRs that reference the issue (via "Closes #N" or the locus/issue-N branch naming convention).

Sprint Mode

Iterate on all open agent PRs that have feedback.

locus iterate --sprint

---

How It Works

1. Fetches comments on the target PR(s).

2. Filters for actionable feedback (excludes bot-generated comments).

3. Passes the feedback context to the AI agent.

4. The agent makes changes on the existing PR branch.

5. Changes are committed and pushed to the PR.

---

Examples

# Iterate on all agent PRs with feedback
locus iterate

# Iterate on a specific PR
locus iterate --pr 15

# Find PR for issue #42 and iterate
locus iterate 42

# Iterate on all sprint PRs with feedback
locus iterate --sprint

# Preview without executing
locus iterate --dry-run

# Iterate on PR 15 with a specific model
locus iterate --pr 15 --model claude-sonnet-4-6