Tov Law Grow Your Autonomous Law Firm Free AI Resources for Lawyers
§ 07 · Automation

Clio MCP: Connect Claude to Clio in 20 Minutes

Clio MCP is the connector that lets Claude read your Clio data: matters, contacts, tasks, and billing. There’s no official version yet, but the free open-source clio-mcp server does the job today, and setup takes about 20 minutes.

Here’s why you want it. Clio knows everything about your caseload. Your matters, your contacts, your tasks, your unbilled time. And your AI knows none of it.

That gap costs you every day. “Which matters have deadlines this week” means logging into Clio, clicking through filters, and scanning a list. “How much unbilled time is sitting on the Ramirez file” means another login, another report. Your practice management system is a database you can only talk to with your mouse.

I fixed this at my own firms with MCPs, the connectors that let Claude read your real systems. I’ve written about what MCPs are and which ones to connect first. This article is the Clio-specific version: one open-source server, one API key, about 20 minutes.

There’s no official Clio connector for Claude yet. What exists instead is clio-mcp, an open-source MCP server built by Yellen Legal that talks to Clio’s API and exposes it to Claude.

Free. Open source. You hold the keys, and you can read every line of what it does before you run it. I connected my own Claude to Clio this way using Claude Code, and the whole thing took one sitting.

Here’s the setup.

Step 1: Get a Clio API key

Log into Clio, then go to the Clio developer portal. Create a new application. Clio will ask for a name and a callback URL; the clio-mcp README tells you exactly what to enter.

When the application is created, Clio gives you credentials. Copy them somewhere safe. Not a sticky note, not a shared doc. A password manager. This key can read your case data, and we’ll come back to what that means in the security section.

Step 2: Install the clio-mcp server

Go to the GitHub repo and follow the README. It’s a clone-and-install, the standard pattern for open-source MCP servers. If the commands look foreign, paste the README into Claude and ask it to walk you through each one. That is not a joke. It’s how half my staff got their first MCP running.

Step 3: Add it to Claude’s config

Claude finds MCP servers through a config file. In Claude Code it’s a one-line add command; in Claude Desktop it’s a JSON file where you paste an entry pointing at the clio-mcp server and your API credentials. The repo README has the exact block to copy.

This is the step where people get nervous because it looks like programming. It isn’t. You’re filling in two blanks: where the server lives and what your credentials are.

Step 4: Restart and test

Restart Claude. Then ask it: “List my open matters.”

If the setup worked, Claude calls Clio and answers. If it didn’t, the error message plus the README plus Claude itself will get you unstuck. Most failures are a typo in the config path or a credential pasted with a trailing space.

What you can actually do once connected

The test query is a party trick. The daily value looks like this:

That last one deserves a pause. Stale matters are where malpractice claims and bar complaints are born, and no partner reviews the full matter list weekly. Claude will, in one sentence, every Monday if you want. That standing Monday check is a small taste of what AI agents do across a whole firm.

A field note from my team

My AI manager got a Clio Manage connector working for us by logging in directly through the site, and Clio Grow is next on his list. In our conversations, Clio has been receptive about making a public connector, so a native one may come. Until then, the open-source route works today, and switching later costs nothing.

One more thing we learned the hard way: if both an MCP and a Zapier connection exist for a tool, connect both. The MCP wins for conversational lookups and judgment calls. Zapier wins for triggered automations that fire at 2 AM without anyone asking. Different actions, different strengths.

Clio MCPZapier + Clio
Best forConversational lookups, ad hoc questionsTriggered automations that run unattended
Runs whenYou askAn event fires
CostFree, open sourceZapier plan
SetupAbout 20 minutes, one API keyAbout 10 minutes per Zap

Security: treat the API key like a client file

Because it is one, functionally. Whoever holds that key can read whatever the key’s scopes allow. So:

None of this is exotic. It’s the same hygiene you’d apply to a Clio login, applied to a key.

20 minutes, then plain English forever

That’s the trade. One setup session, and from then on your practice management system answers questions instead of making you click for them. And once Claude can read the matter, drafting the paperwork from it is the natural next step; that’s legal document automation, which sits in the automation stage of the roadmap.

Your next action: open developers.clio.com, create the application, and grab the key. That’s the only step Claude can’t do for you.

Questions lawyers ask

Is there an official Clio connector for Claude?
Not yet. Clio has been receptive about building a public connector, so a native one may come. Until then, the open-source clio-mcp server by Yellen Legal works today, and because it's open source you can read exactly what it does before you install it.
Is connecting Claude to Clio safe for client data?
The API key is the risk surface, so treat it like a client file. Use the narrowest scopes Clio offers, start read-only, revoke keys when staff leave, and remember your confidentiality duties apply the same as with any vendor that touches case data.
Do I need to know how to code to set this up?
No. You copy commands from the repo README, paste your API credentials into a config file, and restart Claude. If you can install a browser extension, you can do this. Claude Code will even walk you through it if you ask.
Should I use the MCP or a Zapier connection to Clio?
Both, if you have both. The MCP is better for conversational lookups and ad hoc questions. Zapier is better for triggered automations that run without you. They cover different actions, so connecting both costs you nothing and doubles what you can do.

Go deeper