Connecting Claude to your project management with MCP
Give Claude context: tickets, resources and design, all available through an MCP connector.
An AI assistant is useful when it knows the context. Without it, Claude gives generic answers; with it, Claude can summarize a project’s progress or draft a ticket that fits what already exists.
MCP in two sentences
The Model Context Protocol (MCP) is an open standard that lets an assistant call external tools. An MCP server exposes tools; the assistant decides when to use them.
What Grepo exposes
- list_tickets and get_ticket to read tickets.
- create_ticket, update_ticket_status and add_comment to act on the board.
- list_resources, list_repos and list_skills for context.
- get_conception and get_need to read the design as Markdown, save_need to write it.
- report_usage to report the tokens consumed, shown on the ticket.
Plugging in the connector
The connector is personal: a single address, https://mcp.grepo.app/mcp, for all your projects. On claude.ai, add it as a custom connector and sign in to Grepo. For Claude Desktop or Claude Code, create a named key in “My account › MCP connector” and paste the configuration snippet provided.
Each project then decides what it opens up: its admins enable the connector and choose the tools, grouped by service. You can, for example, allow tickets to be read without allowing them to be changed. A key can be revoked in one click, without affecting the others.
Working rules for Claude
A project can also tell Claude how to work on it: turn every request into a ticket, comment at each step, link every ticket to the needs it fulfills… Claude reads these options and instructions before acting, and they apply to everyone who connects it to the project.
A few requests to try
- “Summarize what’s in review on the e-commerce project.”
- “Create a ticket for the pagination bug, linked to the web-front repository.”
- “Which rules apply to the Place an order need?”