Alongside its API, Climbo offers an MCP server for both Agency Mode and Client Mode, so you can connect the platform directly to an AI assistant (LLM). This article explains how it works and how to connect it.
MCP for Any LLM
The Two Addresses
Each set of API docs has its own MCP, authenticated the same way as the API: with your API key. The Agency Mode MCP is at doc.climbo.com/mcp, and the Client Mode MCP is at doc.mysaas.site/mcp. You can connect either to any LLM that supports MCP connectors, and then query the platform in plain language.
How It Actually Works
The MCP doesn't offer one tool per feature. It gives the LLM a set of tools for working with the API documentation: list the available specifications, list the endpoints, search through them, read the details of a specific endpoint, and execute a request. So when you ask a question, the LLM first looks up which call answers it, then makes that call with your API key.
What That Means in Practice
Two things follow from this. First, anything the API can do, the MCP can do — and nothing else. Second, on a broad question the LLM may need a few attempts to find the right endpoint, and a specific question gets a faster, more reliable answer than a vague one. Naming what you want ("list my clients created last month") works better than asking for something open-ended.
Example: Connecting to the Claude Desktop App
Adding the Connector
As an example, here's how to connect the Agency Mode MCP to Claude in the desktop app:
1. Below the chat, click the + button, then go to Connectors > Manage Connectors. 2. Click Add in the top right, then Add custom connector. 3. Give it a name — for example, Climbo — and set the URL to doc.climbo.com/mcp. 4. Add an authorization header named x-api-key, and paste in your Agency Mode API key as its value. 5. Allow the connector to actually use its tools: set Always Allow for both read-only tools and write/delete tools. 6. Confirm the connection.
The MCP is now available in Claude, ready to query.
Other AI Apps
Any tool that supports custom MCP connectors over HTTP works the same way: the URL above, plus the x-api-key header with your key. In Claude Code, for instance, you add it from the command line with `claude mcp add` instead of from a settings panel.
A Word of Caution
The MCP Can Change Things, Not Just Read Them
Because the MCP can execute any documented call, an LLM connected to it can create, change and delete real data — clients included. If you'd rather it only answered questions, don't grant blanket approval for write and delete tools: approve calls one at a time, and read what it proposes before confirming.
AI Chat for Clients
Built In, No Setup
Your clients have an AI chat built directly into their platform, connected for them — no separate LLM setup needed. They can ask it for an in-depth breakdown of last week's reviews, a friendly-toned message template, or to bulk-add a long list of phone numbers — tasks they'd otherwise do manually. See the AI Chat article in Client Mode.
What the Agency Mode MCP Can See
Your Whole Agency
Agency Mode doesn't have a built-in AI chat yet, but you can connect its MCP to any LLM using the URL above. The Agency Mode MCP isn't limited to your own agency: it also has access to your clients' data. You could ask it for a general performance analysis of your clients' reviews, or which clients received fewer than 5 reviews last month. Questions like these help you spot clients who aren't getting real value from the platform — a good early signal of churn risk — so you can reach out and help them before they consider leaving.
MCP Is Only as Capable as the API
If It Isn't Documented, It Can't Be Done
Both MCPs are built on the calls listed in the API docs. If an action isn't documented there, the MCP can't do it either, and you'll need to do it manually for now. So the docs also answer the question "can the AI do this for me?": if the call exists, the MCP can make it. We keep the API documentation growing, and everything we add to it becomes available through the MCP at the same time.