Why this matters
In May 2026, a former Latham & Watkins lawyer named Will Chen released "Mike," an open-source legal AI platform built to rival Harvey and Legora. Open-source means the code is public — anyone can take it, run it, and change it. He named it after Mike Ross from Suits, a wink at Harvey, which is itself named after Harvey Specter. The "OSS" stands for open-source software.
The reaction was extraordinary. Within 72 hours, Mike had over 1,000 GitHub stars and 300 forks — the fastest adoption any legal tech project has seen. Chen says he built the whole thing in about two weeks.
Here is why that should make every managing partner sit up. Harvey is valued at $11 billion. Legora recently raised at around $5.5 billion. Mike does much of the same core work — drafting, reviewing, bulk document review — for free.
So an old assumption just broke. Good legal AI no longer has to cost a fortune. The model that powers Mike — Claude or Gemini — is something you plug in yourself. The expensive part was never really the intelligence. It was the packaging.
But free software is not a free solution. That distinction is the whole point.
How Mike actually works — and where it stops
Strip away the headlines and Mike is a web app that does what a lawyer using AI does. It drafts documents. It reviews them. It searches a vault of files. It runs bulk review across many documents at once. It follows saved workflows. It runs on Claude or Gemini, hosted on Microsoft's GitHub.
The feature that matters most is not really a feature. It is where the software lives.
Mike can run locally. That means firm files never touch an outside server. For a profession built on confidentiality, that is the headline. Chen says some firms have refused to upload sensitive documents to Harvey's hosted vault. With Mike, the firm owns the software outright. There is no third-party vendor holding your clients' secrets.
For an Indian firm, this lands harder still. India's data law, the DPDP regime, is still finalising its rules on moving personal data outside the country. A tool that keeps everything on your own machines sidesteps that whole question. Local deployment is not just a security nicety here. It is a compliance advantage.
Now the honest part. Mike is not a finished replacement.
An independent review by Brightline Labs spent an hour inside the live demo. Their verdict was careful. The parity is real for daily drafting and review. It is partial for firm-wide workflows. It is absent for anything that needs proprietary case-law databases. So Mike is strong at the everyday work and thin at the deep, specialised work.
There is a legal catch too. Mike uses a licence called AGPL v3. In plain terms: if you let outsiders use your version of Mike — say, through a client portal — you may have to share your source code with them. That is fine for internal use. It becomes a real consideration the moment clients touch it.
And the version on mikeoss.com is only a demo. The actual value comes from installing Mike on your own systems. That is not download-and-go. It is an engineering project.
What this means for your firm
This is where most people ask the wrong question. They ask: free or paid?
The real question is: rented or owned?
With Harvey, you rent a finished service. The price includes a deployment team that installs it, secures it, connects it to your documents, and tunes it to your work. With Mike, the software is free — but that deployment work does not vanish. It moves to you.
Someone still has to install it. Secure it. Connect it to your document system. Build your workflows. Keep it running and updated. Handle the licence. None of that lives in the GitHub repository. All of it decides whether Mike becomes a daily tool or an abandoned browser tab.
That someone is a legal engineer. A person who understands both the law and the tooling, who can turn "we downloaded it" into "we use it every morning."
Chen himself argues that lawyers overestimate how hard legal AI is. He is half right. The model is easy now. The deployment is not.
So here is how to decide.
If your firm can put real engineering behind it, Mike is a serious option — especially for data-sensitive Indian work, where keeping files in-house is an advantage, not just a preference.
If it cannot, the free software will sit unused. A paid service that arrives with its own team may turn out cheaper in the end.
Mike did not remove the need for expertise. It relocated it. The firms that win with it will not be the ones who saved the most on licence fees. They will be the ones who had someone to bring it to life.
That is the work we do at aircounsel — the deployment team for the firm that wants to own its AI, not just rent it.