An MCP issue tracker for Claude Code: Linear, Jira, GitHub or something simpler

Once Claude Code writes most of your code, the hard part is keeping track of what it is doing: what is open, what is done, what it found along the way. An issue tracker with an MCP server fixes that. The agent files a ticket when it finds a bug, comments when it is stuck and closes the ticket when the work is done, and you read the board instead of scrolling a terminal.

What an MCP issue tracker does

MCP, the Model Context Protocol, is how an AI agent talks to another program. An issue tracker with an MCP server gives the agent a small set of tools: list the tickets, read one, create one, change its status, add a comment. With those, the same tickets work for you and for the agent.

• The agent remembers between sessions. A new Claude Code session reads the open tickets and knows where things stand.
• You review instead of supervise. The agent writes what it did and why on the ticket; you read it when you have time.
• Nothing gets lost. The bug the agent noticed while fixing something else becomes a ticket, not a line in a chat you'll never find again.

Linear MCP

Linear runs an official remote MCP server. In Claude Code you add it with claude mcp add --transport http linear-server https://mcp.linear.app/mcp, then run /mcp in a session to sign in. It uses OAuth, so there is no API key to paste. Linear's own docs note that remote MCP is still young and a connection sometimes needs a second try.

Good fit: a team that already plans in Linear, with cycles, projects and triage. Less good: one person and an agent who don't need cycles and roadmaps, where Linear's structure is more than the work.

Jira MCP (Atlassian)

Atlassian has an official remote MCP server for Jira and Confluence Cloud, added in Claude as a custom connector with OAuth. It works with Jira Cloud only, not Jira Server or Data Center, and Atlassian publishes rate limits per plan. Check Atlassian's documentation for which Cloud plans include it.

Good fit: a company that already runs on Jira and wants the agent inside the same workflow. Less good: a side project, where setting up Jira is a project of its own.

GitHub Issues

GitHub has an official MCP server too, and Claude Code works well with the gh command line. If your code is on GitHub, issues sit right next to it. Good fit: open source and code-only work. Less good: work that isn't code, and following along from your phone.

A tracker made for one person and an agent

If you build alone or in a small team with Claude Code, you may want less: numbered tickets, a status and a thread on each, nothing else. That is what Pinpal Projects is. It has an MCP server at https://projects.mypinpal.com/mcp (OAuth, works in Claude, Claude Code, ChatGPT and other MCP clients), apps for iPhone, iPad and Mac, and a web app. The agent files, comments, attaches logs and closes tickets; you follow the board on your iPhone, iPad or Mac. It is free for up to 20 tickets, then $1.99 a month for unlimited, with the first week free.

How to choose

• Your team already lives in Linear or Jira: connect that one. The agent should work where the people work.
• All your work is code on GitHub: GitHub Issues and the gh CLI.
• You and an agent, code and other projects mixed: a small tracker you can read anywhere, like Pinpal Projects.

A CLAUDE.md that makes the agent use the tracker

Whichever you pick, tell the agent how to use it. A few lines in your project's CLAUDE.md are enough:

## Tickets
- Work comes from the open tickets in the tracker. Start each session by listing them.
- One ticket at a time. Set it to in progress when you start.
- When you find a new problem, file a ticket for it instead of fixing it on the side.
- When you're done, comment what you changed and how you tested it, then set it to done.
- If you're stuck, set the ticket to blocked with one concrete question.

The last line matters most: a blocked ticket with a clear question is the difference between an agent that waits for you and one that guesses.