MCP Server

The JuhJuh MCP server gives Claude Code and Claude Desktop direct access to your organization. Create tickets, plan sprints, search your knowledge base, and manage projects without leaving the conversation.

MCP (Model Context Protocol) is an open standard for connecting AI assistants to external tools. JuhJuh's MCP server exposes 52 tools, 2 guided prompts, and 2 live resources that Claude can call during any conversation.

How it works¶

sequenceDiagram
    participant C as Claude Code
    participant M as MCP Server
    participant J as JuhJuh API

    C->>M: tool call (e.g. create_ticket)
    M->>J: authenticated API request
    J-->>M: response
    M-->>C: structured result

Claude sends tool calls to the MCP server over a local stdio connection (or HTTP in production). The server authenticates with JuhJuh using your API key, executes the request, and returns the result directly into the conversation.

Setup¶

Claude Code¶

1. Open your project directory
2. Create a .mcp.json file in the project root:

json { "mcpServers": { "juhjuh": { "command": "python", "args": ["-m", "juhjuh_mcp.server"], "env": { "JUHJUH_API_KEY": "jjk_your_api_key_here", "JUHJUH_API_URL": "https://api.juhjuh.com/api/v1" } } } }

1. Restart Claude Code. The server starts automatically.

Claude Desktop¶

1. Open Settings > Developer > Edit Config
2. Add the JuhJuh server to mcpServers:

json { "mcpServers": { "juhjuh": { "command": "python", "args": ["-m", "juhjuh_mcp.server"], "env": { "JUHJUH_API_KEY": "jjk_your_api_key_here", "JUHJUH_API_URL": "https://api.juhjuh.com/api/v1" } } } }

1. Restart Claude Desktop. You should see JuhJuh listed in the MCP servers panel.

Remote connection (production)¶

For hosted environments where you cannot run local processes, connect directly to the production MCP endpoint:

https://mcp.juhjuh.com/mcp/

This uses streamable HTTP transport. Configure your MCP client to connect to this URL with your API key in the Authorization header.

Getting your API key¶

1. Go to Settings > API Keys in your JuhJuh organization
2. Click Create API Key
3. Copy the key (starts with jjk_). Store it securely. You will not see it again.

The key inherits your organization membership and permissions. It can access any project you have access to.

Install¶

bash pip install juhjuh-mcp-server

Verify:

bash python -m juhjuh_mcp.server --help

Tools¶

The MCP server exposes 52 tools organized by domain. Every tool requires an org_slug parameter identifying your organization.

Tickets¶

Projects¶

Sprints and boards¶

Pages and knowledge¶

Knowledge categories: architecture, domain, pattern, reference, learning, glossary, endpoint, model_detail, service, business_rule, workflow, integration, config, page.

Agents¶

Connections¶

Batch operations¶

Intake and triage¶

Marketplace¶

Brand voice¶

Templates¶

Organization¶

Guided prompts¶

Prompts are multi-step workflows that guide Claude through a structured process. They combine multiple tool calls into a coordinated sequence.

Sprint planning¶

Name: sprint_planning Parameters: project_key, org_slug

Walks through five steps:

1. Review active and completed sprints for carry-over items
2. Scan the backlog for unassigned and high-priority tickets
3. Check team capacity by listing members
4. Recommend tickets for the next sprint based on priority and capacity
5. Suggest a sprint goal and duration (1 to 4 weeks)

Ticket triage¶

Name: triage_ticket Parameters: ticket_key, org_slug

Walks through five steps:

1. Retrieve full ticket details
2. Search knowledge base for relevant architecture and pattern context
3. Assess and recommend a priority level with justification
4. Update the ticket with the recommended priority
5. Add a triage summary comment

Resources¶

Resources are read-only data streams that Claude can access for context.

Pagination¶

All list endpoints support pagination with two parameters:

Error handling¶

The MCP server returns structured error messages for common failure modes. Your API key and organization access are validated on every request.

The server retries rate-limited requests automatically with exponential backoff (up to 3 retries).

Environment variables¶

Example conversation¶

Here is what a typical Claude Code session looks like with the MCP server connected.

You: "Create a ticket in ACME for adding dark mode to the settings page"

Claude calls: create_ticket(org_slug="acme-engineering", project_key="ACME", title="Add dark mode to settings page", ticket_type="feature", priority="medium")

Result: Ticket ACME-147 created.

You: "What's in the current sprint?"

Claude calls: list_sprints(project_key="ACME", status="active") then get_sprint(project_key="ACME", sprint_id="...")

Result: Sprint 12 has 8 tickets, 3 in progress, 2 done.

You: "Search the knowledge base for our authentication patterns"

Claude calls: search_knowledge(org_slug="acme-engineering", query="authentication patterns", category="architecture")

Result: 3 entries found covering session management, OAuth flow, and token refresh.

Related¶

• Integrations for connecting Jira, GitHub, Bitbucket, and Slack
• CLI for terminal-based infrastructure management
• Agents for custom AI agents with tools and triggers
• Tickets for the ticket lifecycle from creation to merge
• Projects for organizing repositories and team access
• Permissions for understanding API key access scope