Skip to main content

Introduction to the Craft.io MCP server

Connect an AI assistant to your Craft.io product data using the MCP server.

Written by Maayan Ayalon

AI assistants like Claude, ChatGPT, Cursor, Gemini, and Visual Studio Code are powerful, but they typically operate without access to your live product data.

That means your AI assistant typically does not know:

  • Your roadmap

  • Your OKRs

  • Customer feedback

  • Feature hierarchy

  • Delivery status

  • Dependencies

  • Internal terminology

Without live context, AI output is often generic and disconnected from the way your organization actually works.

The Craft.io MCP server changes that by securely connecting AI tools directly to your Craft.io Workspace, Portfolio and Feedback Portal. Instead of manually copying information into prompts, your AI assistant can now work with real product context in real time.

Want early access? The Craft.io MCP server is currently in private beta. Join the MCP waitlist


What is MCP?

MCP, or Model Context Protocol, is an open standard that allows AI assistants to securely connect to external systems and tools.. Think of it as a universal adapter - one standard that works across Claude, Cursor, VS Code, Gemini, and other AI tools, rather than a separate integration for each one.

Before MCP, teams typically had two choices:

  • Manually paste context into prompts

  • Build retrieval systems that could search documents but could not take action

MCP introduces a third, operational AI model. With MCP, AI assistants can:

  • Read live workspace data

  • Analyze operational context

  • Create and update records

  • Automate workflows securely
    ​


How MCP works

MCP consists of three components:

  • Host - the AI tool you are working in (Claude, ChatGPT, Cursor, VS Code, Gemini)

  • Client - the MCP layer built into the AI tool that handles communication and tool execution.

  • Server - the external system exposing capabilities to the AI assistant (i.e. the Craft.io MCP server).

The Craft.io MCP Server is remote-hosted, so there is nothing to install locally.

Once you have an API token, you'll be able to connect to the Craft.io MCP server at https://mcp.craft.io/mcp, using the AI assistant of your choice.


What you can do

Most AI interactions require you to bring the context. When your AI tool is connected to Craft.io via MCP, your workspace becomes the context instead.

Read and search your workspace

Use natural language to:

  • Find epics, features, OKRs, or feedback

  • Surface blocked or unowned work

  • Analyze dependencies across initiatives

  • Summarize roadmap progress

  • Identify unresolved customer pain points

Instead of manually navigating dashboards, your AI assistant can retrieve and analyze information directly from your workspace.

Create and update items

AI assistants connected through MCP can also perform operational actions, such as:

  • Create epics and user stories

  • Generate feedback entries

  • Update statuses and priorities

  • Assign owners

  • Modify backlog items

  • Generate follow-up work from customer calls or support tickets

This allows AI to move beyond summarization into real workflow execution.

Draft product documentation

Generate:

  • PRDs

  • Release notes

  • Acceptance criteria

  • Sprint summaries

  • Stakeholder updates

Content can be written directly into Craft.io item descriptions without copy-pasting between tools.

Capture feedback automatically

The Craft.io MCP Server can help structure and route customer insights from:

  • Gong transcripts

  • Support tickets

  • Slack conversations

  • Meeting notes

AI assistants can extract themes, identify trends, and create structured feedback entries automatically.

Monitor delivery and reporting

Because the AI has access to live Workspace and Portfolio data, it can help teams:

  • Track roadmap progress

  • Monitor OKR health

  • Identify delivery risks

  • Surface stalled work

  • Generate stakeholder briefings

All reporting is grounded in real-time operational data instead of manually assembled updates.



Security and access

Every MCP connection is tied to a specific Craft.io user and inherits their permissions.

Organizations can:

  • Scope API keys to specific workspaces, portfolios, or feedback portals

  • Configure read-only or read/write access

  • Maintain existing permission boundaries

All AI-driven actions are logged with a full audit trail, including:

  • The originating user

  • The executed operation

  • Affected records

  • Via Agent attribution
    ​

MCP usage is subject to a fair use policy at the account level. At this time, there is no token cost per user.


Getting started

The Craft.io MCP server is currently in closed beta. Want early access? Join the MCP waitlist

Already have access? Connect your AI tool:


What comes next

  • Guru AI overview - Craft.io's built-in AI for your workspace and feedback portal

  • Public API - learn more about the Craft.io public API
    ​

Need more guidance? πŸ™‹ Our LIVE support team (at the bottom right corner of your screen) replies to ANY question!

Did this answer your question?