How to let your AI client work with the content projects your team already runs in Clipflow

Give an AI assistant the context it needs without copying project details into every prompt

Clipflow runs a Model Context Protocol server, so an AI client like Claude can read your workspace directly. Ask what is due this week, or which posts failed to publish, and the answer comes from your actual projects rather than something pasted into a chat.

Endpoint: https://api.clipflow.co/mcp

This is separate from the AI Agents inside Clipflow. Agents live in your workspace and comment in it. The MCP server brings your workspace into an AI client you already use.

---

Before you start

• An AI client that supports remote MCP servers over HTTP.
• A Clipflow account with access to the organizations you want to read.

---

Step 1: Add the server to your client

Add https://api.clipflow.co/mcp as a remote MCP server in your client’s connector settings.

The server publishes its protected-resource metadata for discovery, so most clients handle registration and authorization themselves without any manual configuration:

https://api.clipflow.co/.well-known/oauth-protected-resource/mcp

---

Step 2: Authorize

Your client opens Clipflow, you sign in, and a consent screen lists the scopes being requested, grouped into Analytics, Content, Projects & Media, and Workspace. Approve it and the client holds a token scoped to what you agreed to.

The connection sees what you see. It cannot reach organizations you are not a member of, and it cannot do anything your own permissions do not allow.

---

Step 3: Use it

Once connected, these tools are available:

Tool	What it returns
whoami	The signed-in user
list_organizations	Organizations you belong to
list_projects	Projects, filterable by organization, assignee, channel, status, category, priority, and title
get_project	One project
project_overview	A project with its tasks, media, and posts
list_media_files	Media files, filterable
get_media_file	One media file
list_posts	Posts, filterable
get_post	One post
channel_product_lookup	Channels and the post types they support
analytics_overview	Channel and post performance

Every tool reads. Nothing in this set changes your workspace, so connecting an AI client cannot alter a project, publish a post, or delete a file.

Questions that work well:

• “Which projects are overdue in the agency organization?”
• “Summarise what my team shipped last week.”
• “Which posts failed to publish this month, and on which channels?”
• “What is assigned to Sam right now?”

---

Tips and best practices

• Name the organization when you belong to several. It saves the client a lookup and gets you a straighter answer.
• Approve only the scopes you need. Analytics-only is enough for reporting questions.
• Treat it as read-only. For writes, use the API.
• Revoke connections you have stopped using from Organization settings → API.

---

Frequently asked questions

---

Related articles

• API Overview: OAuth, scopes, and REST endpoints
• Webhooks: push events to your own systems
• AI Agents: AI that works inside the workspace