jtk — Jira Toolkit

The most comprehensive dedicated Jira MCP server in the open-source ecosystem. A dual-mode Go binary that works as both a rich CLI tool and an MCP server for AI agents.

11 MCP tools · 4 prompt templates · Full CLI · Single binary · Zero dependencies

Why jtk?

Related MCP server: ctk

Features

• Dual Mode — CLI for humans, MCP server for AI agents, same binary

• Git Awareness — Auto-detects Jira issue keys from branch names (e.g., feature/PROJ-123-add-login → PROJ-123)

• Dynamic Permission Introspection — Queries /mypermissions at MCP startup, only registers mutation tools your token allows

• Dev Status API — Surfaces branches, PRs, and commits linked to any issue via Jira's 3-step dev-status endpoint

• Issue Metrics — Cycle time, lead time, time-in-status breakdown with status transition history

• MCP Prompts — Standup summary, sprint status, release notes, dev dependency tree

• Token-Efficient — Consolidated action-based tools minimize schema overhead. ResponseFlattener strips bloated JSON. Full ADF↔Markdown conversion

• Agile-First — Boards, sprints, backlogs, sprint mutations, active sprint detection

Installation

# From source
git clone https://github.com/zach-snell/jtk.git && cd jtk
./install.sh  # builds and copies to ~/.local/bin

# Or build manually
go build -o jtk ./cmd/jtk

Pre-built binaries available on the Releases page.

Quick Start

# Authenticate
jtk auth

# Get current issue from git branch
jtk issues get

# Search with JQL
jtk issues search --jql "project = PROJ AND status = 'In Progress'"

# Create an issue
jtk issues create --project PROJ --type Task --summary "Fix login bug"

# Sprint overview
jtk boards list
jtk boards active-sprint --board 1

CLI Commands

jtk auth                    Authenticate with Jira Cloud
jtk issues                  Issue CRUD, search, comments, transitions, links
jtk boards                  Agile boards, sprints, backlogs
jtk projects                List, get, create projects
jtk users                   Search and get users
jtk versions                Project versions/releases
jtk worklogs                Time tracking

MCP Server

Stdio Transport (Claude Desktop, Cursor, OpenCode, etc.)

{
  "mcpServers": {
    "jira": {
      "command": "/path/to/jtk",
      "args": ["mcp"],
      "env": {
        "JIRA_DOMAIN": "your-domain",
        "JIRA_EMAIL": "you@example.com",
        "JIRA_API_TOKEN": "your-api-token"
      }
    }
  }
}

Streamable HTTP Transport

jtk mcp --port 8080

Environment Variables

MCP Tools (11)

MCP Prompts (4)

Security

Three-layer safety model:

1. Token scopes — Atlassian scopes control which APIs the token can call (403 if missing)

2. Permission introspection — jtk queries /mypermissions at startup and dynamically hides mutation tools your account lacks

3. Tool denial — Explicitly hide tools: JIRA_DISABLED_TOOLS="manage_boards,manage_worklogs"

Token scopes — use classic

When creating an API token with scopes, prefer classic scopes for full read + write:

read:jira-user,read:jira-work,write:jira-work,manage:jira-project

(Read-only: read:jira-user,read:jira-work.) jtk auto-detects the token type, so granular scopes also work — but granular is Beta and finicky: "fat" endpoints require every satellite scope (e.g. /myself needs read:user + read:avatar + read:group + read:application-role), and granular personal API tokens currently return 401 Unauthorized; scope does not match on POST requests through the gateway, making them effectively read-only. jtk auth prints copy-paste scope lists for both.

Development

go test -race ./...          # Run tests
golangci-lint run ./...      # Lint
go build -o jtk ./cmd/jtk   # Build

License

Apache 2.0