AI agents use a remote MCP connector at /api/mcp/v1. Staff sign in with their Car Storage Software account, approve scopes, and the agent gets tools for your facility data.

Open AI agents (MCP) under Settings → Integrations. This is separate from the API keys card.

What you will accomplish

• Enable AI agents (MCP) from Settings → Integrations.
• Copy the connector URL from that dialog.
• Add it in Claude, Cursor, or Claude Code.
• Sign in as staff, approve scopes, and confirm the agent can list cars or customers.

Connector URL

https://www.carstoragesoftware.com/api/mcp/v1

Local development: https://css.localhost/api/mcp/v1.

Claude (claude.ai / Desktop)

1. Enable the integration

   Open Settings → Integrations → AI agents (MCP) and turn it Enabled. Copy the connector URL.

2. Add a custom connector

   In Claude, open Customize → Connectors (Team/Enterprise owners: Organization settings → Connectors). Click Add / Add custom connector. If asked for type, choose Web.

3. Paste the URL

   Paste the connector URL. Leave OAuth client ID/secret blank — Claude discovers auth automatically. Click Add.

4. Connect and try it

   Click Connect, sign in as facility staff, approve scopes (start read-only), then ask Claude to list cars or customers.

Cursor

1. Enable the integration

   Open Settings → Integrations → AI agents (MCP) and turn it Enabled. Copy the connector URL.

2. Add the server

   Open Cursor Settings → Tools & MCP, or edit ~/.cursor/mcp.json, and add:

   {
     "mcpServers": {
       "car-storage": {
         "url": "https://www.carstoragesoftware.com/api/mcp/v1"
       }
     }
   }
   

3. Connect and try it

   Click Connect if Cursor shows Needs authentication. Sign in as staff, approve scopes, then ask the agent to list cars or customers.

Claude Code

1. Enable the integration

   Open Settings → Integrations → AI agents (MCP) and turn it Enabled.

2. Add the server

   Run:

   claude mcp add --transport http car-storage https://www.carstoragesoftware.com/api/mcp/v1

3. Authenticate

   In Claude Code, run /mcp, select car-storage, choose Authenticate, sign in as staff, and approve scopes.

Other clients

Paste the same URL into any host that supports remote MCP (Streamable HTTP) with OAuth. You will use the same staff sign-in and consent screen.

Clients that only support local stdio MCP should use Connect the API with a team key instead.

Scopes and safety

OAuth scopes match API key scopes (read:cars, write:customers, …). Consent defaults to read scopes. Grant write only when the agent should create or update records — reconnect the connector after enabling write scopes or the first write will ask you to sign in again.

Revoke access from the AI client, or turn AI agents (MCP) Off for the team.

Intake example (drop-off + photos)

Typical agent sequence for a vehicle drop-off with photos:

1. list_cars — find or confirm the car (avoid duplicates).
2. list_event_types — resolve “Vehicle Drop Off” → value: vehicle_drop_off (check config.requiresParkingSpot).
3. list_locations / list_parking_spots (or zones) — pick where it parked.
4. create_cars_by_id_events — body like { "activityType": "vehicle_drop_off", "parkingSpotId": "…", "occurredAt": "2026-09-29T02:30:00.000Z", "notes": "…" }. Needs write:cars and write:parking.
5. create_cars_by_id_photos — one image per call with sessionId from step 4 and dataBase64 (or sourceUrl). Chat attachments are not uploaded automatically; the agent must pass image bytes/URL into this tool.

Do not use marketing asset upload for car gallery photos, and do not rely on create_appointments alone for intake — that only writes a calendar row.

Platform note (ops)

If Bot Protection or a WAF sits in front of the app, allow Claude’s published egress IPs and do not challenge these paths: /api/mcp/v1, /.well-known/oauth-*, /oauth/*.

Troubleshooting

What to do next

For REST keys and Swagger, see Connect the API.