> ## Documentation Index
> Fetch the complete documentation index at: https://docs.octopoda.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Use Octopoda from Cursor and Claude Code over MCP

> Connect the Octopoda MCP server to Cursor or Claude Code so your AI tool can start scans, check progress, list findings and fetch fix instructions.

The Octopoda CLI includes an MCP server, so AI coding tools such as Cursor and Claude Code can use Octopoda directly: start a scan of the project you are working on, wait for it, read the findings and apply their fixes. The server runs on your machine with `octopoda mcp` and uses the account you signed in to with `octopoda login`.

## Before you start

1. [Install the CLI](https://docs.octopoda.dev/cli/install.md).
2. Sign in: `octopoda login`. The MCP server uses this sign-in, so it has no credentials of its own to configure.

## Connect your AI tool

> [tab] Claude Code
>
> Add the server from your project folder:
>
> ```bash theme={null}
> claude mcp add octopoda -- octopoda mcp
> ```
>
> Then start Claude Code in that folder. Run `/mcp` to check that `octopoda` is connected.

> [tab] Cursor
>
> Add Octopoda to `.cursor/mcp.json` in your project, or to `~/.cursor/mcp.json` to use it in every project:
>
> ```json .cursor/mcp.json theme={null}
> {
>   "mcpServers": {
>     "octopoda": {
>       "command": "octopoda",
>       "args": ["mcp"]
>     }
>   }
> }
> ```
>
> Open Cursor's MCP settings to check that `octopoda` is enabled.

Any other tool that can run a local MCP server over standard input and output can use the same command: `octopoda mcp`.

## Try it

Ask your AI tool in plain language:

```text theme={null}
Scan this project with Octopoda and fix anything critical or high.
```

The tool starts a code scan, checks on it every minute or two until it completes, then reads each finding's fix and applies it. A code scan uses 5 credits and usually takes 10 to 30 minutes.

## Tools

| Tool | What it does | Arguments |
| - | - | - |
| `octopoda_scan_code` | Uploads the project and starts a code scan. Returns right away with the scan ID and link. | `path` (optional, defaults to the current folder), `instructions` (optional guidance) |
| `octopoda_scan_status` | Returns a scan's status, current stage and the findings proven so far. | `scan_id` (required) |
| `octopoda_list_findings` | Lists open findings in your account, most severe first. | `severity` (optional: `critical`, `high`, `medium` or `low`) |
| `octopoda_get_fix` | Returns what is wrong, why it matters and the fix instructions for a finding. | `finding` (required: a key such as `F-142`, or a finding ID) |

`octopoda_scan_code` uploads the same files as `octopoda scan`, so `.env` files and private keys are left out. See [Choose what gets scanned](https://docs.octopoda.dev/scanning/excluding-files.md).

Scans started from your AI tool show **AI tool** as their trigger on the **Scans** page.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.