otter
First run

First 10 minutes

Get from zero to a grounded chat (and an optional approved patch) on your machine. Pick Docker for the full UI, or the CLI if you prefer the terminal.

1. Install

docker pull manasdutta04/otter

Docker: open http://127.0.0.1:3000/app. CLI: run otter after npm i -g @otter-engg/cli (same package for pnpm / yarn / bun).

2. Connect GitHub

In the UI click Connect GitHub, or in the CLI run /login. You install the Otter GitHub App — you do not paste Client secrets. Details: GitHub.

3. Point at a model

Recommended: Ollama on the host — ollama pull qwen2.5-coder:7b. In Docker open http://127.0.0.1:3000/app/models and choose Local Ollama. In the CLI use /model. More: Models.

4. Import a repository

From the workspace, import a GitHub repo (or /import owner/repo in the CLI). Otter builds intelligence — stack, entry points, architecture signals — before any coding.

5. Ask, plan, then approve

  • Chat — ask how something works; answers cite real files.
  • Planner — request a change; review steps and risks.
  • Coding — link the plan, generate a patch, approve, apply, optional PR.
  • CLI: /create … follows the same plan → approve → implement path.

What stays local

Repos, patches, and model traffic stay on your machine (or your chosen LLM endpoint). Otter is self-host first — see Self-host for the full checklist and Docker / CLI for surface details. Package: @otter-engg/cli.