grist-go-mcp

Quentinchampenois/grist-go-mcp

3.2

If you are the rightful owner of grist-go-mcp and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.

Grist MCP Server is a work-in-progress project designed to facilitate the management of organizations, workspaces, and documents using the Model Context Protocol (MCP).

Tools
1
Resources
0
Prompts
0

Grist MCP Server

[⚠️] Work in progress, not ready for production yet.
[⚠️] TODO: Must minimize prompts and responses to limit tokens usage (around 8000 tokens for the sample request)
[⚠️] Disclaimer: This is just a for-fun project, not actively maintained, but happy to share and contributions are welcome !

Functions:

  • ✅ list_orgs : List all organizations
  • ✅ create_workspace: Creates a new workspace in organization
  • ✅ create_new_docs: Creates one or more documents in workspace
  • ✅ create_new_tables: Add new tables with columns to a document
  • ☑️ query_sql: Query SQL (API not yet available)
  • ☑️ add_record: Add a record in a document (API not yet available)
  • ☑️ get_record: Get a record in a document (API not yet available)

Examples

System prompt example available at

Prompt:

Create a personal finance tracker in Grist with the following structure:

  1. Tables needed:
    - Transactions: columns for Date, Description, Category, Amount, Payment Method, and Notes
    - Categories: columns for Category Name, Type (Income/Expense), Budget Limit
    - Monthly Summary: columns for Month, Total Income, Total Expenses, Net Balance

Getting started

  1. Clone the project
$ git clone https://github.com/quentinchampenois/grist-mcp.git 
$ cd grist-mcp
  1. Build the server
$ go mod tidy
$ go build -o grist-mcp-server

💡 Binary grist-mcp-server must be present in $PATH to be callable

🚀 MCP Server is ready to use ! But how ?
If you try to run it directly $ ./grist-mcp-server process will run but nothing happens, and it's the expected behaviour.

Use @modelcontextprotocol/inspector to debug

Package @modelcontextprotocol/inspector starts a WebUI to test directly your MCP server, test output based on provided input.

Requirements

  • Node

How to

Start inspector as following

$ npx @modelcontextprotocol/inspector grist-mcp-server

Access WebUI at http://localhost:6274/?MCP_PROXY_AUTH_TOKEN=<GENERATED_TOKEN>

Configure client using Cherry studio and Scaleway AI provider

In this example we will use Cherry Studio as client and model qwen3-235b-a22b-instruct-2507 from Scaleway provider.

Requirements

  • Configure account on Scaleway and generate API token (~5 minutes)
  • Cherry studio installed

Configure Qwen3-235b from Scaleway on Cherry Studio

Open Cherry Studio and configure the Scaleway provider: (cf: README on another project for full chery studio steps) 💡 Very important, setup the environment variables in your cherry studio MCP configuration :

  • GRIST_ENDPOINT=http://localhost:8484
  • GRIST_API_KEY=<YOUR_API_KEY>, API_KEY from GRIST WebUI

🚀 Once Cherry studio let's give a try with a basic prompt

I want to manage my tennis club within Grist, including: members, available fields and hours of lessons. 

You must create in Grist : 
* 1 new workspace in my Grist organization
* 3 documents in this newly created workspace

Example:

Prompt: cherry_studio_prompt.png

Result: grist_view.png

Extra