MCP server
Face recognition inside your AI assistant
Connect Claude, ChatGPT, Cursor or VS Code to SightRadar. Ask in plain language and the assistant indexes photos, finds a person from a selfie and runs batches, using 39 tools on your account.
Connector URL
https://mcp.sightradar.com/mcpSign-in happens on sightradar.com. Every call costs the normal API price, from your prepaid credits.
you ask
Which photos in wedding-2026-10 is the person in this selfie in?
the assistant calls
list_collectionsfinds wedding-2026-10search_faces93 credits
or try
- Create a collection called wedding-2026-10 and index every photo at these URLs.
- How many credits do I have left, and what did I spend this week?
- Show me the batches that failed yesterday and retry their failed photos.
two minutes
Add SightRadar to your app
Pick your client. Nothing to install: the server runs at mcp.sightradar.com, and the first call opens your browser to sign in.
- 1On claude.ai or in Claude Desktop, open Settings → Connectors → Add custom connector.
- 2Paste https://mcp.sightradar.com/mcp and choose Connect.
Your browser opens sightradar.com to sign in and approve.
- 1Open Settings → Apps & Connectors → Create.
- 2Paste https://mcp.sightradar.com/mcp as the server URL and pick OAuth as the authentication.
ChatGPT sends you to sightradar.com to sign in and approve.
- 1Add the server from your terminal:
claude mcp add --transport http sightradar https://mcp.sightradar.com/mcpThen run /mcp in a session and choose Authenticate.
- 1Add the server to ~/.cursor/mcp.json:
{
"mcpServers": {
"sightradar": { "url": "https://mcp.sightradar.com/mcp" }
}
}Cursor opens your browser to sign in the first time it uses the server.
- 1Add the server to .vscode/mcp.json in your project:
{
"servers": {
"sightradar": { "type": "http", "url": "https://mcp.sightradar.com/mcp" }
}
}VS Code asks you to sign in the first time it starts the server.
Using Windsurf, Cline or another client? Clients that support remote MCP servers with OAuth use the same URL. The guide also shows how to pass an API key header instead.
sign in with SightRadar
Your account, your rules. The app never sees a key.
Sign in on sightradar.com
Your AI app opens sightradar.com in the browser. Sign in with email and password, Google, or an email link, the same as the console.
Choose what it may do
The page names the app, the address it sends access to and the organization it connects to. Untick any permission you don't want. Account access always starts off.
A key just for this app
SightRadar creates an API key named “MCP: <app name>” for the connection and hands it to the MCP server directly. The key never passes through the browser or the app.
Disconnect any time
Every connection is listed under Connected AI apps on the API keys page. Disconnect signs the app out and revokes its key at once.
permissions
39 tools in five groups
You choose the groups when you approve. Tools outside them are not offered to the assistant at all.
- read10 tools
View collections, batches, webhooks, your credit balance and usage.
Free
- search4 tools
Search, detect and compare faces. Stores nothing.
Spends credits
- write8 tools
Create collections, index photos, register selfies, submit batches and add webhooks.
Spends credits
- delete5 tools
Delete and restore collections and photos, and delete webhooks.
Free
- account12 tools
Manage API keys, profile, email alerts, auto-recharge, orders and team.
Can charge card
Off by default. It can change auto-recharge, which can charge your card.
pricing
No MCP fee
A tool call is an API call at the normal price, billed per photo processed, never per face. 100,000 credits = $1. Paid tools take an idempotency key, so a retry is not charged twice.
| Tool | Credits |
|---|---|
| search_faces, detect_faces, index_photo, register_selfie | 93 per photo |
| compare_faces | 93 per image |
| search_by_point_id | 62 per search |
| submit_batch, requeue_failed_batch_photos | 62 per photo |
| Reading, collections, webhooks, deleting | free |
93 credits is $0.00093, 62 credits is $0.00062. See pricing for top-ups and storage.
open source
Prefer to run it locally?
The sightradar-mcp npm package runs the same tools on your machine with an API key. It can also read image files from folders you allow. Needs Node.js 20 or newer.
npx -y sightradar-mcp{
"mcpServers": {
"sightradar": {
"command": "npx",
"args": ["-y", "sightradar-mcp"],
"env": { "SIGHTRADAR_API_KEY": "frs_your_key" }
}
}
}Cursor, Windsurf, Cline and most other clients accept the same block.
questions
Before you connect
- Does the AI app see my API key?
- No. The key is created on sightradar.com and sent to the MCP server over a server-to-server channel. The app receives OAuth tokens that only work against mcp.sightradar.com.
- Is using it through MCP more expensive?
- No. Every tool call is a normal API call at the normal per-photo price, billed from your prepaid credits. There is no MCP fee.
- Can I limit what the assistant can do?
- Yes. Tick only the permission groups you want when you approve. A connection with only Read cannot spend credits or change data. Tools outside the granted groups are not offered to the assistant at all.
- Can it read photos from my computer?
- The hosted server takes image URLs or base64 data. The local server can also read image files from folders you allow with SIGHTRADAR_FILE_ROOTS.
A selfie is biometric data. Register selfies only with the person's consent, as the Acceptable Use Policy requires.
Connect your assistant now
No account yet? Signing in from your AI app lets you create one. New accounts start with free credits.