Now you can use any agent to add new features, reschedule an initiative, or start building off a product specification.
When your daily work spans multiple agents, progress stalls at the boundaries they can’t cross. You’ll know you’ve hit one when you’re endlessly copy-pasting, ferrying markdown files between tools, or doing manual data entry at the tail end of an otherwise AI-driven task.
The Productboard MCP server bridges those divides by allowing you to employ the AI tools and agents you’re already using to access and update the product data, specifications, and documents in your Productboard workspace. That includes standard LLM tools like Claude and ChatGPT, coding agents like Claude Code, Codex, or Cursor, and AI prototyping tools like Lovable or Claude Design.
Whether you’re automating a prioritization workflow, looking for new ways to get data into Productboard, or updating Productboard data on the go from your favorite AI tool’s mobile app, the MCP server makes it possible.
Create and update features, subfeatures, initiatives, releases, objectives, key results, companies, users, feedback, documents, and more directly from your agent, without leaving your workflow.
Update custom drivers and other data fields on features and initiatives (like value, risk, impact, confidence, and effort) to automate prioritization workflows.
Add features to initiatives, move an initiative to a new release, or create key results beneath an objective. Agents can see how Productboard entities are related and make updates.
Access your latest product specifications from coding agents and AI prototyping tools to initiate the product delivery process. Use your agents to make updates and leave comments back on the original specifications in Productboard.
MCP (Model Context Protocol) is a shared standard that lets AI agents connect to other tools. The Productboard MCP server is the connection point on our side. It gives your agent a menu of tools it can use to work with your Productboard data, including information on how your workspace is organized, so it doesn’t have to figure that out from scratch. Authorize once from your agent of choice, and from then on you’ll be able to prompt it in plain language to access or update data on your behalf in your Productboard workspace.
Read on for quick start directions. For additional information, see the following support articles: Connecting a coding agent to Productboard via MCP and Create and manage entities with an AI agent.
Admins start by enabling the MCP Server at the workspace level.
Enable the MCP server:
Admins can revoke every connection workspace-wide with Disconnect all members.
Each workspace member connects to the Productboard MCP server from their own AI tools and agents. Once authorized, they will only be able to access what’s allowed by their workspace role and permissions.
Steps:
The following URL is used for the configuration: https://mcp.productboard.com.
If prompted to specify which type, choose HTTP. For authorization, choose OAuth.
The final step is authenticating with a Productboard login.
Members can disconnect the MCP connector at any time.
With Claude Code CLI already installed, run the following in your terminal:
claude mcp add --transport http productboard https://mcp.productboard.comWith Codex CLI already installed, run the following in your terminal:
codex mcp add productboard --url https://mcp.productboard.comAdd the following to your MCP configuration JSON file:
{
"mcpServers": {
"productboard": {
"type": "http",
"url": "https://mcp.productboard.com"
}
}
}
