Skip to main content

CheckFlow MCP Server

The CheckFlow MCP server lets an AI assistant — Claude, or any other client that speaks the Model Context Protocol — work in your CheckFlow workspace. Connect it once, and you can ask your assistant in plain language to start a checklist, tell you what is overdue, fill in a task, build a template from a written procedure or tidy up a Data Set, and it does so through the same operations as the REST API.

note

The MCP server is available on every plan, as is the REST API.

What You Can Ask For​

The server offers 147 tools, grouped into families. A few examples of what an assistant can do with them:

You askThe assistant
"What's on my plate this week?"Reads your Tasks grid, filtered to tasks due this week.
"Start an Invoice Review for INV-2041, due Friday, and assign the approval to Sarah."Creates the checklist from the Invoice Review template with its parameters, sets the due date and assigns the task.
"Mark the bank details check as done and note that the IBAN was verified by phone."Fills in the task's fields, adds a comment and completes the task.
"Which supplier onboardings have been stuck for more than two weeks?"Searches checklists and reads their activity.
"Turn this procedure document into a template."Reads the authoring guide, composes a template document, validates it, fixes what the validator reports and publishes it.
"Add these twelve suppliers to the Suppliers Data Set."Creates the records in one bulk call.
"Set up a checklist every Monday at 9 am for the weekly stock check."Creates a schedule.
Tool familyToolsCovers
Checklists15Listing, searching, creating, completing, archiving, sharing and tagging checklists.
Checklist Tasks27Completing tasks, fields and tables, files, assignees, due dates, comments, snoozes, tags.
Standalone Tasks26Raising and managing tasks outside checklists, with sub-tasks, comments and files.
Workspace, People and Tasks Grid12Members, groups, tags, the Tasks grid and its saved views.
Template Authoring30The workspace summary, the schema and authoring guide, templates, versions, permissions, upgrades and drafts.
Data Sets21Data Sets, fields, records, views and CSV import and export.
Schedules9Recurring schedules and their runs.
Webhooks7Webhook subscriptions, deliveries and replays.

The server also publishes one resource, the template authoring guide, which a client can attach to a conversation. See How the MCP Server Works.

Who the Assistant Acts As​

An assistant connects in one of two ways, and each decides who it works as:

  • Signed in with your CheckFlow account. The assistant works as you, in the workspace you were signed in to when you connected. There is nothing to choose: a connection made by signing in always acts as the person who signed in, never as anybody else or as the workspace. Administrators, and Members with the AI.Assistant permission, can connect this way; Guests cannot. See Signing In.
  • With an API key. Every key acts as one member of the workspace — or as the workspace itself — chosen when the key is created.

Either way, your assistant works with that member's identity and access. It sees what they can see, its changes are recorded in their name, and it is refused where the API refuses them — see Who a Key Acts As for the permission checks that apply.

For a personal assistant, sign in, or use a key that acts as you. That way "my tasks" means your tasks, snoozes hide tasks from your own Tasks grid, and the activity feed shows your name against the changes you asked for. A key that acts as the workspace cannot use the tools about a person, such as the Tasks grid and snoozing.

Keeping It Safe​

A connected AI assistant can make real changes to your workspace, and a model can misunderstand a request. Before you connect one:

  • Sign in where your client allows it. There is then no key to store or leak, and the connection can do no more than you can.
  • Use a key that acts as the least powerful member who can do the job. Everything the assistant does is limited to what that member could do in the app, and managing webhooks needs an Administrator. A key that acts as the workspace is not limited by member permissions but cannot use the tools about a person, so read Who a Key Acts As before you choose. Avoid giving an assistant an Administrator's key unless it needs one.
  • Give each assistant its own key, named for it on the API & Integrations tab, so that you can see when it was last used and revoke it on its own. Assistants connected by signing in are listed there too, under Connected AI assistants.
  • Keep your client's confirmation prompts on. Every tool tells the client whether it only reads, whether it changes something and whether the change is destructive. Twenty-six tools are marked destructive — among them deleting a checklist, a standalone task, a Data Set or its records, removing a comment, file or table row, replacing a Data Set's records from CSV, publishing a new template version and upgrading running checklists — and most clients ask before running them. Every other tool that writes is marked non-destructive, so a client that honours the hints does not ask about completing a task or filling in a field.
  • Treat the key as a secret. Store it in your client's secret or environment settings rather than in a shared configuration file.
  • Disconnect what you no longer use. Revoke a key on the API & Integrations tab, or disconnect an assistant you signed in to on the AI Assistants tab of your profile. See Disconnecting an Assistant.
warning

Deleting a checklist, a standalone task, a Data Set or its records through the MCP server is as permanent as doing it in the app. Read what the assistant proposes before you approve a destructive tool.

Requirements​

  • An MCP client that supports the Streamable HTTP transport — Claude on the web, desktop and mobile, Claude Code, ChatGPT, Cursor, VS Code and many others. Microsoft 365 Copilot agents cannot connect yet. See Microsoft 365 Copilot.
  • Either a CheckFlow account to sign in with, or an API key, created by an Administrator on the API & Integrations tab of Team Management.
  • The AI.Assistant permission for the member the assistant acts as, unless they are an Administrator. New Members are given it. See Choose Who Can Connect an AI Assistant.
  • Allow AI assistants to connect switched on for the workspace. It is on unless an Administrator has turned it off on the AI tab.

See Connecting a Client.

Server Addresses​

RegionMCP endpoint
United Stateshttps://api.checkflow.io/mcp
Europehttps://api-eu.checkflow.io/mcp

Use the address for the region your workspace is in. A request to the other region's server is refused with WRONG_REGION, and the message names the address to use instead. If you connected by signing in, remove the connection and connect again with that address.

Last updated on