MCP Overview
GitScrum MCP Server provides AI assistants with full operational access to your project management stack through the Model Context Protocol.
Open Source — GitScrum MCP Server is open source under the MIT license. Available on npm and GitHub. Model Context Protocol server for GitScrum — Claude, GitHub Copilot, Cursor, and any MCP-compatible client full operational access to your project management stack.
GitScrum MCP Server connects AI assistants to your GitScrum workspace through the Model Context Protocol. It exposes 29 tools with 160+ operations across 6 categories, giving your AI assistant the same operational capabilities your team uses daily in the GitScrum web application — task management, sprint planning, time tracking, CRM, analytics, and more.
Everything your team does in the GitScrum web app, your AI assistant can now do through natural language conversation.
What is the Model Context Protocol?
The Model Context Protocol (MCP) is an open standard that enables AI assistants to interact with external tools and data sources in a structured, secure way. Instead of copying and pasting data between your project management tool and your AI assistant, MCP creates a direct bridge.
With the GitScrum MCP Server running, your AI assistant can:
- Query your workspaces, projects, and tasks in real time
- Create and update tasks, sprints, user stories, and epics
- Track time, generate standup reports, and review analytics
- Manage clients, invoices, and proposals through ClientFlow CRM
- Search across your entire project management stack
All of this happens through natural language. You ask your AI assistant a question or give it an instruction, and it calls the appropriate GitScrum MCP tools behind the scenes.
Why Use MCP for Project Management
Traditional project management requires constant context switching — opening browser tabs, navigating dashboards, filling forms. MCP eliminates that friction by letting your AI assistant handle the operational work while you stay focused on what matters.
Natural language operations. Instead of navigating through menus to create a task, say "Create a high-priority bug in the Auth project for the login timeout issue" and the AI assistant handles field mapping, project selection, and task creation.
Cross-tool intelligence. Your AI assistant can correlate data across tools. Ask "Which projects are over budget and have overdue tasks?" and it queries budget tracking, task filters, and sprint data in a single conversation.
Automation without scripts. Complex workflows that would require API scripting become conversational. "Create a sprint for next week, add the top 5 backlog items, and assign them to the frontend team" executes multiple operations sequentially.
Team-wide visibility. Any team member with an MCP-compatible client can interact with the same project data. Developers, managers, and stakeholders all benefit from conversational access to the project management stack.
Tool Categories
The GitScrum MCP Server organizes its 29 tools into six categories. Each tool uses a consolidated action parameter, reducing LLM context tokens by approximately 80% compared to individual tool definitions.
Core
The foundation tools for daily project management operations.
| Tool | Key Actions | Purpose |
|---|---|---|
task | my today get create update complete subtasks filter by_code duplicate move notifications | Full task lifecycle management |
sprint | list all get kpis create update stats reports progress metrics | Sprint planning and tracking |
workspace | list get | Workspace navigation |
project | list get stats tasks workflows types efforts labels members | Project configuration and data |
time | active start stop logs analytics team reports productivity timeline | Time tracking and productivity |
Planning
Tools for backlog grooming, story mapping, and workflow configuration.
| Tool | Key Actions | Purpose |
|---|---|---|
user_story | list get create update all | User story management |
epic | list create update | Epic-level planning |
label | list create update attach detach toggle | Label taxonomy |
task_type | list create update assign | Task type configuration |
workflow | create update | Kanban workflow stages |
Collaboration
Tools for team communication, documentation, and knowledge management.
| Tool | Key Actions | Purpose |
|---|---|---|
discussion | all channels channel messages send search unread markread createchannel update_channel | Team messaging and channels |
comment | list add update | Task and item comments |
wiki | list get create update search | Project documentation |
note | list get create update share revisions | Personal and shared notes |
note_folder | list create update move | Note organization |
search | — | Full-text search across all data |
ClientFlow CRM
Client relationship management, invoicing, and proposal tools.
| Tool | Key Actions | Purpose |
|---|---|---|
client | list get create update contacts interactions add_interaction | Client management |
invoice | list get stats create update issue send mark_paid | Invoice lifecycle |
proposal | list get stats create update send approve reject convert | Proposal management |
clientflow_dashboard | 8 reports | CRM analytics and reporting |
clientflowcrossworkspace | 4 reports | Cross-workspace CRM data |
Insights PRO
Advanced analytics, standup automation, and budget monitoring.
| Tool | Key Actions | Purpose |
|---|---|---|
standup | summary completed blockers team stuck digest contributors | Automated standup reports |
analytics | 10 reports | Manager dashboard analytics |
activity | feed userfeed notifications activities taskworkflow | Activity feed and audit trail |
budget | projectsatrisk overview consumption burn_down alerts events | Budget tracking and alerts |
Authentication
Secure session management tools.
| Tool | Description |
|---|---|
auth_login | Initiate OAuth 2.0 device code flow |
auth_complete | Complete browser-based authorization |
auth_status | Check current session and token validity |
auth_logout | Clear stored credentials locally |
Supported Clients
The GitScrum MCP Server works with any client that implements the Model Context Protocol. Tested and verified clients:
| Client | Platform | Configuration |
|---|---|---|
| Claude Desktop | macOS, Windows | claudedesktopconfig.json |
| GitHub Copilot | VS Code | .vscode/mcp.json |
| Cursor | macOS, Windows, Linux | Cursor MCP settings |
| Windsurf | macOS, Windows, Linux | Windsurf MCP settings |
| Continue | VS Code, JetBrains | Continue configuration |
All clients use the same MCP Server binary. The only difference is the configuration file location and format for each client.
Security Model
The GitScrum MCP Server is designed around the principle of least privilege with multiple layers of protection.
| Layer | Protection |
|---|---|
| Operations | Only CREATE, READ, UPDATE. DELETE is blocked at both MCP and API layers. |
| Authentication | OAuth 2.0 Device Grant (RFC 8628) — credentials never touch the MCP server. |
| Transport | HTTPS-only communication with TLS 1.2 minimum. |
| Token storage | Local filesystem (~/.gitscrum/auth.json) with restricted permissions. |
| Rate limiting | Automatic lockout after failed authentication attempts. |
No DELETE operations. This is a deliberate design choice. AI assistants should not be able to permanently destroy project data. All destructive operations must be performed through the GitScrum Studio web application, where human confirmation is required.
Example Conversations
Here are practical examples of what you can accomplish through natural language:
You: "What's on my plate today?"
AI: Fetches your tasks due today across all projects.
You: "Create a sprint for next week with the top 5 backlog items"
AI: Creates the sprint, assigns tasks, and sets the timeline.
You: "Show me which projects are over budget"
AI: Returns burn-down data and flags at-risk projects.
You: "Send the Q1 proposal to Acme Corp"
AI: Creates the proposal, attaches the client, and sends it.
You: "What did the team ship this week?"
AI: Generates a standup digest with completed work and blockers.Architecture
The MCP Server runs as a local process on your machine. It communicates with AI clients via the Model Context Protocol (stdio transport) and with GitScrum's API over HTTPS.
┌────────────────┐ MCP (stdio) ┌──────────────┐ HTTPS ┌─────────────┐
│ AI Assistant │ ◄──────────────────► │ MCP Server │ ◄───────────► │ GitScrum API │
│ (Claude, etc) │ │ (local) │ │ (cloud) │
└────────────────┘ └──────────────┘ └─────────────┘- AI Assistant sends tool calls through the MCP protocol
- MCP Server validates requests, manages authentication, and forwards to the GitScrum API
- GitScrum API processes the request with full access control and returns structured data
The server is stateless between sessions except for the authentication token stored locally. No project data is cached or stored on disk.
Requirements
| Requirement | Minimum |
|---|---|
| Node.js | 18.0.0 or higher |
| npm | 8.0.0 or higher |
| GitScrum account | Free or paid plan |
| MCP client | Any MCP-compatible client |
Next Steps
- Quick Start: Install and authenticate in under 5 minutes.
- Configuration: Detailed setup for Claude Desktop, GitHub Copilot, Cursor, and other clients.
- Authentication: Deep dive into the OAuth 2.0 Device Grant flow.
- Security: Enterprise security model, error handling, and best practices.