FoxCueConnect Claude or Cursor to FoxCue with MCP
Give AI assistants like Claude and Cursor access to your FoxCue sources, ideas, scripts and outputs through the local stdio MCP server and a scoped API key.
FoxCue includes an MCP (Model Context Protocol) server, so AI assistants like Claude Desktop or Cursor can work with your FoxCue sources, video ideas, scripts and outputs. The assistant can search your workspace, suggest ideas from an article, write and revise scripts, and check on renders, all with the permissions of a scoped API key.
The FoxCue MCP server runs locally over stdio. Your MCP client starts it on your own computer from a FoxCue build, and it talks to the FoxCue API with your workspace API key.
What you need
- A local FoxCue build that includes the compiled MCP entrypoint (
backend/dist/developers/mcp.js) and Node.js to run it. - A workspace API key. See Create an API key.
- An MCP client that supports stdio servers, for example Claude Desktop or Cursor.
Connect an assistant
Create a scoped key
In Settings → Developers, create a key with only the scopes your workflow needs. For a read-only assistant, that's sources:read, strategy:read and scripts:read. Add scripts:write and generation:write if it should write scripts.
Copy the client configuration
In Settings → Developers, open the MCP tab. Under Connect an Assistant, copy the MCP client configuration with its copy button, and paste it into your client's private settings. It runs the server with node and sets the environment variables below.
Add your key
Replace <your workspace API key> with your key in the client's secret settings, never in a shared file.
Restart the client
Restart the assistant. FoxCue's tools appear in its tool list.
Environment variables
The MCP server reads these variables. Use the names exactly as written. These variable names may change in a future release.
| Variable | Value |
|---|---|
THREADFLIX_API_URL | The API base URL from Settings → Developers, for example https://app.foxcue.com/api/v1 |
THREADFLIX_API_KEY | Your workspace API key |
THREADFLIX_MCP_PROFILE | workflow (recommended) or legacy |
"env": {
"THREADFLIX_API_URL": "https://app.foxcue.com/api/v1",
"THREADFLIX_API_KEY": "<your workspace API key>",
"THREADFLIX_MCP_PROFILE": "workflow"
}If the server can't start, check that THREADFLIX_API_URL is set and THREADFLIX_API_KEY holds a valid workspace API key.
Tool profiles
- Workflow (recommended, and the profile in the copied configuration): a focused set of tools with typed inputs and structured results. See MCP tools reference.
- Legacy: the older, more granular tools, kept for existing integrations. If
THREADFLIX_MCP_PROFILEisn't set, the server uses this profile.
Try it
Ask your assistant:
- "Search my FoxCue workspace for anything about recurring jobs."
- "Suggest three video ideas from https://www.example.com/blog/post for the HVAC silo."
- "Write the script for the idea I just accepted."
The assistant shows you each tool call. Ideas, scripts and outputs it creates appear in the FoxCue app too.
Security
- The assistant has exactly the scopes of the key you give it.
generation:writelets it start paid AI work. Grant it deliberately.- Revoke the key in Settings → Developers to cut off access immediately.
Frequently asked questions
Can I connect ChatGPT or another web-based assistant?
No. FoxCue's MCP server is local only. It works with assistants that can start a local stdio server on your computer.
Does the assistant see my internal sources?
Only what the key's scopes allow. Internal source text is still kept out of publishable scripts.