@fictura/mcp

Read your Fictura app from any AI coding agent — its live paywalls, funnel, experiments, users, errors, deep-link config and the full documentation — so the agent writes integration code against your real setup instead of guessing.

Everything it exposes is read-only.

Most clients don't need this package

Fictura's MCP server is hosted. If your client supports HTTP servers — Claude Code, Cursor, VS Code, Windsurf — point it straight at the URL:

claude mcp add --transport http fictura https://api.fictura.co/mcp --header "Authorization: Bearer fk_mcp_YOUR_KEY"

This package is a stdio bridge for clients that only speak stdio, such as Claude Desktop.

Get a key

Fictura dashboard → Setup → Coding agents (MCP) → create a key. It is shown once.

Without a key the server still answers documentation questions; your app's data needs one.

Claude Desktop

In claude_desktop_config.json:

{
  "mcpServers": {
    "fictura": {
      "command": "npx",
      "args": ["-y", "@fictura/mcp"],
      "env": { "FICTURA_MCP_KEY": "fk_mcp_YOUR_KEY" }
    }
  }
}

| Variable | Purpose | | --- | --- | | FICTURA_MCP_KEY | Your MCP key. Omit it for documentation-only access. | | FICTURA_API_BASE | Point at another Fictura server. Defaults to https://api.fictura.co. |

What the agent can do with it

Ask it to "add a Fictura paywall to the onboarding flow" and it will pull your real SDK key, check which placement names actually exist, and read the current live config before writing anything. Ask it to "fix the crash in production" and it can read the grouped issue with its stack traces.

Full tool list and per-client setup: https://docs.fictura.co/connections/outgoing/mcp-server

License

MIT