Getting started

Quickstart

Fix your first bug end to end, and read the artefacts it leaves behind.

neo takes a repository and a bug report in plain words. It plans a fix, edits inside a Docker sandbox, and runs your real test suite. It reports success only when the target test passes and the full suite shows no regressions.

Run a fix

export MY_KEY=...   # your OpenAI-compatible router key

neo fix \
  --repo ./your-project \
  --issue "test_mean fails: mean() returns the sum, not the average" \
  --provider openai \
  --model <model> \
  --api-key $MY_KEY \
  --api-base <base-url>

Read what it produced

On a verified fix, neo writes several artefacts under the task's log directory. The rationale is the one worth reading first — it states what was wrong, what changed, and how it was verified, grounded in the actual trace rather than generated after the fact.

Artefact	What it holds
rationale.md	What was wrong, what changed, how it was verified
git.json	Branch name, commit SHA, and a PR description
state.json	Plan, completed steps, files touched, decisions
trace.jsonl	Every prompt, response, tool call and result

Inspect progress

neo status --task-id <task_id>
neo dashboard --logs-dir logs/