Mike wants to be the open-source answer to Harvey and Legora, two VC-backed legal AI platforms signing up law firms at a steady clip. The project handles contract work and document analysis in a codebase you can self-host and audit.
The pitch is straightforward. Bring your own Claude or Gemini API key. Deploy on your own infrastructure. Pay only for model costs, not per-seat licenses that climb every year. For law firms nervous about sending client data to yet another startup's servers, this matters.
Mike ships with a chat interface that reads documents and cites verbatim. There are matter-scoped project workspaces that maintain context across files. The tabular review feature extracts data from hundreds of documents in parallel, with citations back to specific pages.
But early developer feedback suggests the codebase is young. That's not surprising for a new open-source project, but it means firms comparing Mike to battle-tested incumbents should temper expectations. There's also the privilege question. Attorney-client privilege remains a live concern with any AI tool that sends data to external model providers. Self-hosting helps, but it doesn't eliminate the problem if prompts still flow to Anthropic or Google servers.
The real test will be whether Mike can attract enough contributors to mature quickly. Open source works when a community forms around a framework. And legal AI demands accuracy above all else.