Mike is an open-source legal AI platform with document chat, contract drafting, and bulk extraction — deployable on your own infra with your own API keys.
A project called Mike has launched as an open-source alternative to Harvey and Legora, two dominant enterprise legal AI platforms. It offers document-aware chat with verbatim citations, parallel spreadsheet-style extraction across hundreds of documents, reusable workflow templates, and end-to-end contract drafting. Firms bring their own Claude or Gemini API keys, pay only model costs, and can self-host the entire stack — meaning documents never leave the firm's perimeter. It's available as a hosted product or via direct repo clone.
Mike exposes its full prompt-engineering logic, citation parsing pipeline, and data flow in an open codebase — meaning developers can study how a production legal AI is actually built, not just guessing from black-box APIs. You can fork it, wire in your own DMS connectors, swap the citation engine, or extend the extraction layer with custom logic. It uses Claude and Gemini via standard API keys, so integration is straightforward for any team already working with those SDKs.
Clone the Mike repo this week, run it locally against a sample contract using your own Anthropic API key, and benchmark citation accuracy against your current legal document parsing setup to decide whether it replaces a custom build.
Suggested exercise · not independently tested. Access requirements and costs may vary.
Go to the Mike GitHub repo (search 'Mike legal AI open source' or find via the project's landing page) and clone it: git clone [repo-url]
Add your Anthropic API key to the .env file and run the setup script per the README, then upload a sample NDA or credit agreement via the local UI
Ask: 'List all representations and warranties with verbatim quotes and page numbers' and inspect the citation output for hallucinations or broken references
Tags
Sources