# Set up Jot with an AI agent

> Connect Claude Code, Cursor, Codex, Windsurf or any MCP client to Jot (usejot) in two lines, then ask it what is open.

Source: https://docs.usejot.co/docs/ai/setup

Jot (usejot) is click-to-pin bug reporting for live apps. Once your coding
agent is connected, it reads each report with its screenshot, selector and
replay summary, fixes the code, and replies on the reviewer's pin.

<Callout>
  MCP agents are on every plan, Free included. Connecting needs a Jot workspace and, for the CLI, a
  workspace admin to approve it in the browser.
</Callout>

## The lines to paste

### Claude Code plugin (recommended)

```text
/plugin marketplace add usejot/claude-plugins
/plugin install jot@usejot
```

Then ask Claude to connect Jot (it opens your browser), or paste an agent token when Claude Code asks.

### Cursor, Codex, Windsurf, any agent

```bash
npx @usejot/cli login
npx @usejot/cli init
```

`login` opens your browser to pick a workspace. `init` connects the repo. No token to paste.

### By hand

```json
{
  "mcpServers": {
    "jot": {
      "command": "npx",
      "args": [
        "-y",
        "@usejot/mcp"
      ],
      "env": {
        "JOT_API_TOKEN": "<your token>"
      }
    }
  }
}
```

Any MCP client: run `npx -y @usejot/mcp` with JOT\_API\_TOKEN set to an agent token from Settings → Agents.

### claude.ai and ChatGPT

Coming soon.

Agents reply and resolve; who approves is your workspace's choice (by default, the person who reported it).

## Per agent

The plugin or the CLI above does the work. These are only where each agent
keeps its MCP servers, and how to check it worked. Full blocks for each are in
[MCP for coding agents](/docs/ai/mcp#set-up-your-agent).

| Agent         | Where the server goes                                                 | Check it worked                                 |
| ------------- | --------------------------------------------------------------------- | ----------------------------------------------- |
| Claude Code   | the plugin brings it; or `claude mcp add --scope user jot -- jot mcp` | `/mcp` shows `jot` connected                    |
| Cursor        | `.cursor/mcp.json` in the repo, or `~/.cursor/mcp.json`               | Cursor Settings → MCP lists `jot`               |
| Codex         | `codex mcp add jot -- jot mcp`, or `~/.codex/config.toml`             | `codex mcp list` shows `jot`                    |
| Windsurf      | `~/.codeium/windsurf/mcp_config.json`                                 | Cascade's MCP panel lists `jot` after a refresh |
| Anything else | the JSON under **By hand** above                                      | `jot mcp doctor`                                |

Then ask: **"what's open in Jot?"**

## Asking an agent to set it up for you

An agent can run the setup itself. Paste this into it, in the repo you want
connected:

```text
Set up Jot in this repo: run `npx @usejot/cli login`, wait while I pick the
workspace in the browser, then run `npx @usejot/cli init` and tell me which
files it wrote. Then list what's open in Jot.
```

`login` is the one step the agent cannot do for you: it opens your browser so
you choose the workspace. No token is pasted anywhere, and the files `init`
writes hold none, so they are safe to commit.

## Next

* [How agents should use Jot](/docs/ai/agents): the playbook every agent follows.
* [The jot CLI](/docs/ai/cli): every command, including `jot loop`.
* Per-agent pages on the site: [Claude Code](https://usejot.co/for/claude-code),
  [Cursor](https://usejot.co/for/cursor), [Codex](https://usejot.co/for/codex),
  [Windsurf](https://usejot.co/for/windsurf).
