# MCP introduction

> How AI assistants will connect to ReelsRadar via Model Context Protocol.

# MCP introduction

ReelsRadar is building an MCP server so assistants (Cursor, Claude, ChatGPT) can query tracked videos and start tracking jobs with natural language.

## Planned endpoint

```
https://reelsradar.io/api/mcp
```

Transport: Streamable HTTP. Auth: Bearer workspace token (dashboard-created).

## Planned tool groups

- `videos` — search, list, get metrics
- `accounts` — list, pause, analytics summary
- `track` — track URL / bulk track
- `collections` — list and attach items
- `overview` — KPI snapshot
- `export` — request CSV

## Safety rules

- Destructive tools require `confirm: true`
- Never ask users to paste secrets into chat
- Respect plan video limits and rate ceilings

## Status

MCP runtime is on the roadmap. Until it ships, agents should use:

1. [https://reelsradar.io/llms.txt](https://reelsradar.io/llms.txt)
2. [https://reelsradar.io/openapi.json](https://reelsradar.io/openapi.json)
3. Session-authenticated REST from the user's browser automation / local scripts