Self-host on Docker
Docker docs
New ✨ MCP on PyPI — pip install otter-mcp
Easily understand & change code
$ npm i -g @otter-engg/cliDocker for the full local UI, or @otter-engg/cli for the interactive CLI. Repos and data stay on your machine. First 10 minutes
Understand repositories locally
Import a GitHub repo and get architecture signals, entry points, and stack detection — before anyone writes a line of code.
First 10 minutes
Import · repo list · status
Connect a free local model
Point Otter at Ollama on your machine (or any OpenAI-compatible endpoint). One Models page — used for chat, explain, and coding.
Learn more
Providers · discovery · test
Ask the codebase
Grounded chat cites real files. Plan changes with risks and verification steps before any patch is proposed.
Learn more
Grounded answers · citations
Approval-gated coding
Generate patches, review diffs, approve or reject, then open a PR. Otter never silently overwrites your source tree.
Learn more
Diff · approve · PR
How it runs
From install to PR tonight
One path on your machine — no cloud agent, no silent writes.
- Install
Docker for the full UI, or the npm CLI in your terminal.
- Connect GitHub
Install the Otter GitHub App — no Client secrets on your machine.
- Point at a model
Local Ollama (recommended) or any OpenAI-compatible endpoint.
- Import a repo
Otter scans architecture, entry points, and stack before coding.
- Plan → approve → ship
Generate a patch, review the diff, approve, then open a PR.
Ship with Otter in Fast.
Understand first · architecture before patches
Human approval · nothing lands without a yes
Web · CLI · MCP · one brain across surfaces
The details
We're making engineering intelligence accessible
As AI coding tools rush to write files first, Otter starts with understanding. Run it on your machine, bring your own model, and keep approval before every write.
Find out more
Or install the npm CLI
CLI docs
Ready to run Otter?
docker pull manasdutta04/otter