> ## Documentation Index
> Fetch the complete documentation index at: https://docs.edges.run/llms.txt
> Use this file to discover all available pages before exploring further.
# Documentation MCP
> Connect this documentation to Claude, Cursor or VS Code so your assistant answers from the current docs instead of guessing.
These docs are served over the **Model Context Protocol**. Point an AI assistant at the endpoint and it can search and read every page here, so its answers come from the current documentation rather than from whatever it happened to learn.
```
https://docs.edges.run/mcp
```
This is a **documentation** server. It reads these pages and nothing else — it does not call the Edges API, does not run actions, and never sees your API key. To teach an agent how to *call* the API, use [Agent Skills](/v1/guides/agent-skills).
## What it can do
| Tool | What it does |
| ----------------- | ----------------------------------------------------------------------------------------------- |
| **Search** | Finds relevant passages across the documentation and returns them with links |
| **Browse** | Navigates the docs as a filesystem, so an assistant can read a whole page rather than a snippet |
| **Send feedback** | Reports a page that is wrong, out of date or unclear straight back to us |
The feedback tool is the useful one to know about. If your assistant works from these docs and hits something that doesn't add up, it can tell us — that reaches the same place a support message would, and it is the fastest way to get a page fixed.
## Connecting it
Open **Settings → Connectors**, choose **Add custom connector**, and enter:
```
https://docs.edges.run/mcp
```
In a conversation, attach the connector from the attachments menu.
Press `Cmd + Shift + P`, search for **Open MCP settings**, and add the server to `mcp.json`:
```json theme={null}
{
"mcpServers": {
"Edges Docs": {
"url": "https://docs.edges.run/mcp"
}
}
}
```
Create `.vscode/mcp.json` in your project:
```json theme={null}
{
"servers": {
"Edges Docs": {
"type": "http",
"url": "https://docs.edges.run/mcp"
}
}
}
```
No key, no account, nothing to enable on your side — the documentation is public and so is the server.
## When it helps
It is worth connecting when you are **writing code against the API** and want the assistant to check a parameter, a limit or an error label rather than invent one. Questions it answers well:
* Which limit does this action consume, and what is the ceiling?
* What does this `error_label` mean, and how long should I wait?
* Which execution mode should I use for this volume?
* What changed in the API last month?
It does **not** know anything about your workspace — your runs, your identities, your credit balance. Those live behind your API key and are read through the API itself.
## MCP or agent skills?
Both connect Edges to an AI tool, and they do different jobs.
Your assistant can **look things up**. Best when you are asking questions and reading as you build.
Your agent knows **how to call each endpoint** — inputs, URL formats, common mistakes. Best when it is writing the integration.
Using both is the normal setup: skills for how to call the API, MCP for everything else on this site.
## Related
Per-endpoint instructions for coding agents.
The official clients.
First call, in a few minutes.