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

# MCP server

> Connect Claude, Cursor, or any MCP client to Filecheck — check, fix, validate, and optimize print files from an AI agent.

`filecheck-mcp` is a [Model Context Protocol](https://modelcontextprotocol.io) server for the Filecheck API. Once connected, an AI agent can preflight a customer file, apply automatic fixes, validate PDF/A conformance, render page previews, and browse your profiles and rules — with your account's tenancy and billing applied throughout.

There are two ways to run it. Both use the same tools and the same secret key.

<Warning>
  The secret key (`sk_…`) grants full API access to your account. Keep it in your MCP client's local config or server-side environment — never in a repository or a browser.
</Warning>

## Local server (recommended)

The local server runs on your machine via `npx`, so it can upload files straight from disk — file bytes go to Filecheck over a presigned upload, never through the model's context.

```bash theme={null}
claude mcp add filecheck -e FILECHECK_API_KEY=sk_... -- npx -y filecheck-mcp
```

For Claude Desktop or Cursor, add it to the MCP configuration file:

```json theme={null}
{
  "mcpServers": {
    "filecheck": {
      "command": "npx",
      "args": ["-y", "filecheck-mcp"],
      "env": { "FILECHECK_API_KEY": "sk_..." }
    }
  }
}
```

## Remote server

If installing packages is not an option, connect to the hosted endpoint instead. It exposes the same tools minus local-file access — sources are URLs or previously uploaded `fileRef`s.

```bash theme={null}
claude mcp add --transport http filecheck https://api.filecheck.io/mcp \
  --header "Authorization: Bearer sk_..."
```

## Tools

| Tool              | What it does                                                                             |
| ----------------- | ---------------------------------------------------------------------------------------- |
| `check_file`      | Preflight a file against a profile; returns verdict, findings, and a recommended autofix |
| `fix_file`        | Preflight, apply safe automatic fixes, re-check; returns the fixed file                  |
| `validate_pdf`    | PDF/A and PDF/UA conformance via veraPDF                                                 |
| `optimize_file`   | Downsample, recompress, and slim a PDF or image                                          |
| `render_previews` | Render page preview images, so the agent can look at a file                              |
| `get_job`         | Poll a job and re-mint fresh download URLs                                               |
| `list_jobs`       | Recent jobs as lean envelopes                                                            |
| `delete_job`      | Soft-delete a job and its artifacts                                                      |
| `list_catalog`    | Browse profiles, rules, workflows, connectors, and optimize presets                      |
| `get_report`      | Drill into the full preflight report, section by section                                 |
| `save_artifact`   | Save a deliverable to a local path (local server only)                                   |

## Reading results

Every job carries two independent fields, and agents are instructed to read both:

* **`status`** — did the check *run*: `pending`, `running`, `done`, `skipped`, or `error`.
* **`outcome`** — the *verdict*: `pass`, `warn`, `fail`, `fixed`, `unchanged`, `na`, or `mixed`.

`status: done` with `outcome: fail` means the check completed and the file is **not** print-ready. The same [two-axis model](/docs/server/verify-jobs) applies across the whole API.

<Note>
  Download URLs are presigned and expire after about 60 minutes. `get_job` returns freshly signed URLs on every call, so agents re-fetch instead of storing links. Long-running jobs return `pending: true` with polling guidance rather than blocking.
</Note>

The package is MIT licensed and developed at [github.com/PrintApp/filecheck-integrations](https://github.com/PrintApp/filecheck-integrations).