jiramcp

Give your AI assistant full JIRA access without burning half its context window on tool selection.

Most JIRA MCPs expose too many tools. Every time the model picks one, it spends tokens deciding between jira_get_issue, jira_fetch_issue, jira_issue_get...

jiramcp has three:

| Tool | What it does | |---|---| | jira_read | Fetch issues by key, search by JQL, list projects/boards/sprints | | jira_write | Create, update, delete, transition, comment — with dry_run support | | jira_schema | Discover fields, transitions, and allowed values |

Less tool surface area means more of the context window goes to your actual work. The model makes fewer wrong choices, calls fewer redundant tools, and gets to the answer faster. The three tools compose naturally: schema to discover, read to find, write to change.

Your API token never leaves your machine. jiramcp runs as a local process over stdio — no server to host, no proxy in the middle, no credentials sent anywhere except directly to Atlassian.

Prerequisites: Go 1.21+ installed, a JIRA Cloud account.

1. Go to API token management
2. Click "Create API token"
3. Give it a label, click "Create"
4. Copy the token immediately — you can't see it again

claude mcp add-json jira \
'{"command":"go","args":["run","github.com/mmatczuk/jiramcp@latest"],"env":{\
"JIRA_URL":"https://yourcompany.atlassian.net",\
"JIRA_EMAIL":"you@company.com",\
"JIRA_API_TOKEN":"your-api-token"\
}}'

Ask Claude: *"List my JIRA projects"* — if you see your projects, you're good.

Use the same Go binary and env vars. The server speaks standard MCP over stdio.

| | jiramcp | mcp-atlassian | |---|---|---| | Tools | 3 | 72 | | Runtime | Go binary | Python | | Scope |