Quickstart
From an installed elanous to a first change made for you — three commands if you already have a ChatGPT subscription and are signed in to GitHub.
1. See what is missing
elanous doctor
doctor is a report, not a gate: it exits 0 even with nothing configured. For each credential it says whether it resolves, where it came from, what it unlocks, and whether a free fallback exists. Its readiness section names anything that would trip you up next, with one command each; elanous doctor --fix shows the repairs it can make for you, and --fix --yes applies them.
2. Sign in to a model
The main path is a subscription, not an API key:
elanous login openai-codex # ChatGPT device-code sign-in
elanous login status # which providers have tokens
# with the default llm.provider=auto, a ChatGPT sign-in is used first
Other choices (xAI Grok, OpenRouter for Kimi / GLM / Qwen, Anthropic, Gemini, a local model) are in providers.
3. Ask for a change
Go to the project you want changed and say what you want in one sentence:
cd ~/my-project
elanous harness say "add a Usage section to the README with the three commands a new user runs"
What happens:
- elanous writes a goal from your sentence,
- works on it in a separate git worktree (your working tree is not touched),
- runs the project's tests for the files it changed,
- reviews its own change and reworks it if the review finds problems,
- finishes with a pull request (when the repository has a GitHub remote and
ghis signed in) or with a branch you can merge yourself (no remote).
Add --no-auto-merge if you want to merge pull requests yourself. Add --dry-run to see the plan without starting.
4. Talk to it directly
elanous ask "what does src/app.ts do?" # one question, one answer
elanous agent "why is the build failing?" # one turn with file and shell tools
elanous # the terminal UI
Where to look next
- commands — the commands you will actually use
- configuration — where settings live and how to change them
- troubleshooting — messages that look like one thing and mean another