Connecting AI Assistants to Manifestly (MCP Server)

What is the Manifestly MCP Server?

The Manifestly MCP (Model Context Protocol) server lets you connect AI assistants — like Claude, ChatGPT, and other MCP-compatible clients — directly to your Manifestly account. Once connected, you can manage your workflows and runs using natural language conversations.

With the MCP server, your AI assistant can:

Check your workload and show overdue assignments
Start workflow runs by name
Complete, skip, or assign steps
Fill in data collection fields
Browse workflows, runs, departments, and users
Add comments to steps
Manage webhooks

Supported AI Clients

The MCP server works with any AI client that supports the Model Context Protocol, including:

Claude (claude.ai) — Anthropic's AI assistant, via the Connectors feature
Claude Code — Anthropic's CLI tool for developers
ChatGPT — OpenAI's assistant, via MCP connector support
Other MCP-compatible tools and IDE extensions

Setting Up the Connection

Claude (claude.ai)

Open Claude at claude.ai
Go to Settings and look for the Integrations or Connectors section
Open the Manifestly connector in the Claude directory and click to add it
You'll be redirected to Manifestly to log in and authorize the connection
Once you see the "Connected!" confirmation, return to Claude and start chatting

Claude Code (CLI)

Add the Manifestly MCP server to your Claude Code configuration:

Open Claude Code settings (or your .claude/settings.json file)
Add Manifestly as an MCP server with the URL: https://mcp.manifest.ly/mcp
When you first use a Manifestly tool, you'll be prompted to authorize via your browser

ChatGPT

In ChatGPT, go to Settings > Connectors
Add a new MCP connector with the URL: https://mcp.manifest.ly/mcp
Complete the OAuth authorization flow when prompted

What You Can Do

Once connected, ask your AI assistant to read or update Manifestly in natural language. It can:

Summarize your overdue assignments, late runs, and in-progress work
Browse, inspect, export, validate, create, and import workflows
Read and update runs, including fields and closing runs
Complete, uncomplete, skip, unskip, start, stop, approve, reject, or retract run steps
Complete or uncomplete subtasks, assign steps by ID, name, or email, and change due dates
Add comments and read comments on run steps
Create and manage workflow steps, content blocks, and data fields
Browse and update data tables, columns, and rows
Browse departments, roles, workflows, runs, users, and assignments, subject to your Manifestly permissions
Read run source emails and download their attachments when available
Manage webhooks, holiday calendars, holidays, workflow business-day rules, and workflow holiday calendars

Examples include “What needs my attention?”, “Start the daily store opening checklist,” “Assign the next step to Jordan,” “Approve the inspection step,” “Show me the comments on this step,” “Create a workflow step,” “List our data tables,” or “List our webhooks.” The assistant acts as your authenticated Manifestly user, so actions and results follow your existing roles and permissions.

Available Tools Reference

The MCP server currently provides 78 tools, organized into four groups. Tool availability can change as the server evolves; the connected client receives the current definitions.

Smart and information tools

Get My Status, Assign Run Step by Email, Assign Run Step by Name, and Start Run by Workflow Name
About Manifestly and Verify Authentication

Reading data

Departments and roles: list or get departments and list roles
Workflows: list/get workflows, export a workflow, list workflow steps, and get the workflow import schema
Runs and steps: list/get runs, get a run step, list step comments, and list assignments
Fields: get all fields or one field
Users: list or get users (subject to permissions)
Webhooks: list webhooks
Data tables: list/get tables, columns, and rows
Holiday calendars: list calendars and holidays
Run source email: retrieve the source email or a specific attachment

Runs, steps, workflows, and fields

Runs: create, update, set field values, and close runs
Run steps: complete, uncomplete, skip, unskip, start, stop, approve, reject, retract, assign, update due dates, and add comments
Subtasks: complete or uncomplete subtasks
Run fields: set field values; fields can also be updated individually
Workflows: create or import workflows, and validate an import before applying it
Workflow steps: create, update, or delete steps
Step content blocks: create, update, or delete content blocks
Step data fields: create, update, or delete data settings

Webhooks and business calendars

Webhooks: create, update, delete, or unsubscribe webhooks by URL
Holiday calendars: create, rename, or delete calendars; generate regional holidays; add, update, or remove holidays
Workflow scheduling: set a workflow’s holiday calendar and business-day rule

You stay in control

The assistant acts as you, and can only do what your own account permits. It cannot see a department you cannot see, or edit a workflow you could not edit yourself. Every request is authenticated, and you can cut access at any time by disconnecting the assistant or deleting the key it uses.

Two ways to connect

Sign in (OAuth), the recommended route where your assistant supports it. You click Connect, sign in to Manifestly and approve access. There is no key to copy or store, and you can revoke it later from your account.
API key, for assistants that ask for a key instead of a login. Create a key in Manifestly and paste it in when asked.

Which API keys work with MCP

This is the one that catches people out. Only keys in the current mf_1_ format are accepted by the MCP server. Older 32-character keys still work for the REST API, but the MCP server rejects them outright, so a connection attempt with one fails to authenticate rather than failing helpfully.

In Settings → Integrations, an eligible key shows an MCP badge under "Enabled for". If your key does not have it, create a new one for MCP and delete the old key once nothing is using it. See Managing API Keys.

Need Help?

If you're having trouble with the MCP server connection, contact our support team through the help widget in the bottom-right corner of any Manifestly page, or email support@manifest.ly.