Skip to main content
Glama

pancake-api-mcp

A Model Context Protocol (MCP) server that serves the Pancake API documentation to your AI assistant.

Point Claude Desktop, Cursor, or any MCP client at this server and your AI can read the Pancake docs on its own — search endpoints, pull a single endpoint's full contract (parameters, request body, response schema with every $ref expanded), inspect webhook payloads, and read the authentication / rate-limit / webhook-setup guides — without you copy-pasting anything.

The docs are bundled as a static snapshot of the official OpenAPI spec, so documentation lookups are self-contained and offline.

Optionally, configure your access tokens in the server's environment and the call_endpoint tool can also call the live Pancake API for you — the right token (access_token or page_access_token) is picked from each endpoint's OpenAPI security scheme and attached automatically. Without tokens the server stays fully offline and documentation-only.

What's inside

  • 31 REST endpoints across Pages, Conversations, Messages, Statistics, Customers, Posts, Tags, Users, Call logs, Export Data, Page's Contents, and Chat Plugin.

  • 5 webhook events: messaging, conversation, post, subscription, connect_status — each with its full payload schema and an example.

  • 11 guides: authentication & tokens, rate limits, the recommended API usage flow, webhook setup, event types, suspension rules, and best practices.

Related MCP server: Dedalus MCP Documentation Server

Tools

Tool

Input

Returns

search_docs

query, area? (endpoint | webhook | guide | schema | all), limit?

Ranked matches with type, id, and a snippet — start here to discover things

list_endpoints

tag?, keyword?

REST endpoints grouped by tag (id, method, path, summary)

get_endpoint

id (e.g. "GET /pages/{page_id}/conversations")

Full endpoint contract as markdown

list_webhooks

—

All webhook events with summaries

get_webhook

event (e.g. messaging)

Webhook payload schema + example

list_guides

—

Guide section ids + titles

get_guide

id (e.g. authentication)

Full guide section

auth_status

—

Which tokens are configured (masked), plus the other env settings

call_endpoint

id, params?, body?

Live call to the Pancake API — status + response body

Every endpoint, webhook, and guide is also exposed as an MCP resource (pancake-api://endpoint/..., pancake-api://webhook/..., pancake-api://guide/...) for clients that browse resources.

Install

Requires Node.js 18+.

git clone https://github.com/<your-org>/pancake-api-mcp.git
cd pancake-api-mcp
npm install
npm run build

This produces dist/index.js, the server entry point.

Configure your access tokens (optional)

Pancake uses two token types, both passed as a query parameter. Set them as environment variables of the MCP server and call_endpoint will attach the right one per endpoint — you never paste a token into a chat, and tokens are never accepted as tool input.

Variable

Query param

Needed for

How to get it

PANCAKE_USER_ACCESS_TOKEN

access_token

2 account-level endpoints under https://pages.fm/api/v1 (list pages, generate a page token)

Pancake → Account → Personal settings → API Access Token. Valid up to 90 days.

PANCAKE_PAGE_ACCESS_TOKEN

page_access_token

24 page-level endpoints under https://pages.fm/api/public_api/v1|v2 (conversations, messages, statistics, customers…)

Page → Settings → Tools, or POST /pages/{page_id}/generate_page_access_token. Does not expire.

Optional extras:

Variable

Default

Effect

PANCAKE_PAGE_ID

—

Default value for page_id, so you don't repeat it on every call

PANCAKE_READ_ONLY

false

When true, call_endpoint accepts GET only and refuses every write

PANCAKE_TIMEOUT_MS

30000

Timeout for live API calls

Ask your assistant to run auth_status to verify the setup — it reports which variables are set without revealing their values.

Writes are enabled by default. With a page token configured, call_endpoint can really send messages to customers, change tags, and delete data. Set PANCAKE_READ_ONLY=true if you only want the assistant to read. Tokens are read from the environment only, redacted from every tool response and error message, and never written to logs.

Connect it to your AI client

Claude Desktop

Edit your claude_desktop_config.json:

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

  • Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "pancake-api": {
      "command": "node",
      "args": ["/absolute/path/to/pancake-api-mcp/dist/index.js"],
      "env": {
        "PANCAKE_USER_ACCESS_TOKEN": "your-user-access-token",
        "PANCAKE_PAGE_ACCESS_TOKEN": "your-page-access-token",
        "PANCAKE_PAGE_ID": "optional-default-page-id"
      }
    }
  }
}

Drop the env block entirely to run in offline documentation-only mode.

Restart Claude Desktop. You should see the pancake-api tools appear.

Cursor

In Settings → MCP → Add new global MCP server (or edit ~/.cursor/mcp.json):

{
  "mcpServers": {
    "pancake-api": {
      "command": "node",
      "args": ["/absolute/path/to/pancake-api-mcp/dist/index.js"],
      "env": {
        "PANCAKE_PAGE_ACCESS_TOKEN": "your-page-access-token"
      }
    }
  }
}

Replace /absolute/path/to/pancake-api-mcp with the real path where you cloned this repo.

Try it

Once connected, ask your AI things like:

  • "Using the Pancake docs, how do I authenticate?"

  • "What endpoint lists conversations for a page, and what parameters does it take?"

  • "Show me the payload of the messaging webhook."

  • "Search the Pancake docs for anything about tags."

And, with tokens configured:

  • "Check my Pancake token setup." (auth_status)

  • "List the last 60 conversations of page 12345." (call_endpoint)

  • "Send 'thanks for your order' to conversation X." (a write — confirm before you say yes)

Development

npm run build     # compile TypeScript to dist/
npm run dev       # tsc --watch
npm run inspect   # open the MCP Inspector against the built server
npm run sync-spec # (maintainer) refresh spec/ from ../pancake_api_doc/openapi

Updating the bundled docs

The OpenAPI snapshot lives in spec/openapi.yaml and spec/webhook.yaml. A maintainer refreshes them from the source docs project with:

npm run sync-spec                        # assumes ../pancake_api_doc alongside this repo
npm run sync-spec -- /path/to/pancake_api_doc   # or pass the docs repo explicitly
npm run build

License

MIT — see LICENSE.

Available Tools

7 tools
get_endpointA

Get the full contract of a single Pancake REST API endpoint as markdown.

Includes the base URL(s), authentication, all parameters, the request body schema, and the response schema — with $ref schemas expanded into a readable field list — plus any examples. Pass an id from search_docs or list_endpoints.

Inputs: id (required).

get_guideA

Get the full text of a single Pancake API guide section.

Returns narrative documentation such as authentication (how to obtain and use access tokens), rate limits, the recommended integration flow, or webhook setup. Pass an id from list_guides or search_docs.

Inputs: id (required).

get_webhookA

Get the full payload schema and an example for a single Pancake webhook event.

Returns the event's description, the payload structure ($ref schemas expanded into a readable field list), and a sample JSON body. Pass an event name from list_webhooks or search_docs.

Inputs: event (required).

list_endpointsA

List all Pancake REST API endpoints, grouped by tag.

Returns each endpoint's id (e.g. "GET /pages/{page_id}/conversations"), method, path, and summary. Pass the id to get_endpoint for the full contract (parameters, request body, response schema, examples). Optionally filter by tag or keyword.

Inputs: tag, keyword.

list_guidesA

List the Pancake API guide sections (narrative documentation).

These cover concepts rather than a single endpoint: authentication and tokens, rate limits, the recommended API usage flow, webhook setup, event types, suspension rules, and best practices. Pass a guide id to get_guide for the full text.

Inputs: no named parameters.

list_webhooksA

List all Pancake webhook events.

Returns each event's name and summary. Pass an event name to get_webhook for the full payload schema and an example. Webhooks are HTTP POST notifications Pancake sends to your registered endpoint in real time.

Inputs: no named parameters.

search_docsA

Search the entire Pancake API documentation and return ranked matches.

This is the entry point for exploring the docs: search first, then fetch full detail with get_endpoint, get_webhook, or get_guide using the returned id.

Searches across REST endpoints, webhook events, guide sections (authentication, rate limits, setup...), and component schemas. Each result includes its type, the id to pass to the matching get_* tool, and a short snippet.

Inputs: query (required), area, limit.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.6/5.0

    Scored across 7 tools

    Disambiguation5/5

    Each tool has a clearly distinct purpose: search_docs is the cross-cutting entry point, list_* tools provide overviews, and get_* tools retrieve specific details for endpoints, webhooks, and guides. There is no overlap or ambiguity between them.

    Naming Consistency5/5

    All tool names follow the consistent verb_noun pattern: search_docs, list_endpoints, get_endpoint, list_webhooks, get_webhook, list_guides, get_guide. The verbs (search, list, get) and nouns are uniformly applied across the three resource types.

    Tool Count5/5

    Seven tools is well-scoped for a documentation explorer. Each tool serves a necessary role in the search/list/detail workflow, and there are no redundant or missing operations.

    Completeness5/5

    The tool set fully covers the stated purpose of exploring Pancake API documentation: search across all content, list all endpoints/webhooks/guides, and retrieve full details for each. There are no obvious gaps or dead ends in the workflow.

    Related MCP Connectors

    Related MCP Servers