# Connect Claude

Add Qaldy to claude.ai or Claude Desktop as a custom connector, sign in to Qaldy in the window that opens, and ask Claude about your workspace.

URL: https://docs.qaldy.com/mcp/claude/

This works on claude.ai and in Claude Desktop with the same steps. You do not install anything and you do not copy a token.

:::steps
1. Open **Settings**, then **Connectors**, then **Add custom connector**.
2. Name it `Qaldy` and paste the address:

   ```text
   https://mcp.qaldy.com/mcp
   ```
3. Add it. A Qaldy window opens. **Sign in** if you are not, then choose the **workspace** and whether the connector may only **read** or may **read and write**.
4. Press **Allow**. The window closes and Qaldy appears in your connectors.
5. In a chat, turn the connector on from the tools menu.
:::

:::note Menu names move
Claude's settings differ a little between plans and change over time. On Team and Enterprise plans an organization admin may need to enable custom connectors first. If a label here is not where you look, search for "custom connector" in Claude's own help.
:::

![Qaldy's Connect an agent page, with Claude chosen: the steps for a custom connector and the address to paste.](https://docs.qaldy.com/screens/set-up-claude.webp "The Connect an agent page in the console has these steps, with your workspace's address in them. Sample workspace.")

## What you will see in Qaldy

The sign-in page tells you three things, and you should read them:

- **Which app is asking**, and where your browser goes next. A name an app chooses for itself is never trusted. An app Qaldy cannot tell from its address is marked **Not verified**, and an app on your own computer says so.
- **Which workspace.** One workspace per connection. To use a second workspace, add the connector again and choose the other.
- **What it may do.** Read only offers the read tools and nothing else. Read and write adds the tools that change things; a build or an approval still waits for you.

![The Qaldy sign-in for a connector Claude started: recognised from its address, one workspace, read only or read and write, and what it may do.](https://docs.qaldy.com/screens/oauth-consent.webp "The window Qaldy opens. Claude is recognised from its address. Sample request: nothing is connected.")

![The same sign-in for an app Qaldy cannot verify: a Not verified label, a warning that anyone can register an app under any name, and a choice of workspace.](https://docs.qaldy.com/screens/oauth-unverified.webp "An app that only chose the name Claude is marked Not verified, with a warning. Sample request.")

## Try it

Ask in plain words:

- "List my Qaldy apps and tell me which ones are failing."
- "Why did the last run of Support Triage fail?" Claude reads the run step by step, including the call the gate refused.
- "Plan an app that posts to Telegram when coffee beans drop below 5 kg." Claude plans it and shows every permission as a sentence. Nothing is built. When you tell it to approve, Claude asks **you** to confirm in its own window, and the exact plan it showed is what gets built.

## Ending access

Open **Agents** in the console ([https://app.qaldy.com/agents](https://app.qaldy.com/agents)). **Apps signed in** lists every connection with when it last called. End one and it stops working at once. The tokens behind it belong to that app, that workspace and that access, and nothing else.

Next: [Tools](https://docs.qaldy.com/mcp/tools/), [How an agent stays safe](https://docs.qaldy.com/mcp/safety/).