The MCP Bridge: Connecting LLMs to Your Active Project Command Hub
An AI assistant that cannot see your project is a very articulate stranger. MCP is the bridge that lets any compatible assistant work on the real plan, with the permissions you give it and no more.
MCP (Model Context Protocol) is a standard way for AI assistants to discover and call tools. CoMng.AI exposes tasks, notes, dependencies and project health over MCP at https://comng.ai/app/api/external/mcp, authenticated with a scoped API key, so assistants and agents can work in the live project.
- MCP in one sentence: a standard way for AI clients to discover and call tools, so a model sees your project as a set of ready-to-use functions.
- One endpoint: https://comng.ai/app/api/external/mcp, authenticated with an X-API-Key header.
- Same rules as REST: the key's exact permissions and rate limits apply; MCP grants nothing extra.
- Works with the clients you use: Claude, Claude Code, Codex CLI, Gemini CLI, Google Antigravity and any MCP-compatible client.
The Integration Problem MCP Solves
Before MCP, connecting an assistant to a tool meant custom glue: hand-written HTTP calls, bespoke schemas, one integration per assistant. The assistant could not even discover what the tool could do; you had to describe every endpoint in a prompt.
The Model Context Protocol standardises that conversation. A server publishes a list of tools with names and schemas, a client connects, and the model can call them directly. Write the integration once; every compatible assistant can use it.
What an Agent Sees in CoMng.AI
Bind the MCP server and the agent gets the platform's API as tools instead of endpoints to hand-write. Typical tools include tasks_list, task_create, task_update, note_update, activity_create, list_task_dependencies, list_milestone_dependencies and create_milestone_dependency, plus search across tasks, notes, team and budget, and the project health and change request tools.
Because the tools sit on the same API as the product, an agent can read the dependency graph and milestone links that power blocker detection, check what is blocked, and record what it did.
Security: Your Key, Your Rules
MCP uses the same API keys as the REST API. An Admin creates a key and decides what it can do:
- Scope: specific projects or all projects.
- Actions: per-action checkboxes (list, create, update, delete) per resource: tasks, subtasks, notes, activities, team, budget, reports and search.
- Limits: per-key rate limits, optional expiry date and an IP whitelist.
The key is shown once at creation and stored masked afterward. For the governance model around AI actions, see AI Governance and Safety Gates.
Four Things People Do With the Bridge
Ask questions from your assistant
In Claude Desktop or Claude.ai, ask what is blocked this week or what the project health is, and get an answer from live data.
Let a coding agent pick up work
Claude Code or Codex CLI reads the task and its acceptance criteria, works, updates status and logs activity.
Automate hygiene
A scheduled agent with a narrow key can summarise changes or draft notes on a cadence.
Bridge to other tools
Feed project status into the assistant you already use for writing and analysis.
MCP vs REST: When Each Fits
| REST API | MCP server | |
|---|---|---|
| Best for | Scripts, ETL, scheduled jobs, third-party platforms | AI assistants and coding agents |
| Discovery | Read the documentation | The client discovers tools automatically |
| Auth | X-API-Key | The same X-API-Key (or Bearer) carrying the same permissions |
| Limits | Per-key rate limit and permissions | The same, enforced identically |
Setup steps per client are in MCP Server for Project Management; a practical agent workflow is in Claude Code Project Management.
- Custom glue for every assistant
- The model cannot discover capabilities
- Broad credentials shared with scripts
- One endpoint for every compatible client
- Tools are self-describing
- A scoped key per agent, revocable any time
About CoMng.AI
CoMng.AI is an AI-native project management platform that generates a project's plan, tasks, risks and budget from a description or document, monitors project health in real time, and gives both human teams and AI coding agents live access to the same project data, with a person approving every AI-generated change.
Give your AI agents a live plan and a safe way to work in it
Create a scoped API key, connect your agent over MCP and let it pick up work from the same project your team uses.
Frequently Asked Questions
What is MCP (Model Context Protocol)?
An open standard that lets AI clients discover and call tools exposed by a server. For project management it means an assistant can read and update live project data through a defined set of functions.
What is the CoMng.AI MCP server URL?
https://comng.ai/app/api/external/mcp. Authenticate with an X-API-Key header (or Authorization: Bearer) carrying a key from API Management.
Does MCP give an agent more access than the REST API?
No. The MCP endpoint enforces the key's exact permissions and rate limits and grants nothing extra.
Who can create API keys?
Admins. Keys can be scoped to specific projects and actions, rate limited, given an expiry date and restricted to IP addresses.
Which AI clients work with it?
Claude (Claude.ai and Claude Desktop connectors), Claude Code, OpenAI Codex CLI, Gemini CLI, Google Antigravity and any MCP-compatible client.
Related reading
See the full guide: AI Agents, ARI & MCP Guide