Claude & AI Assistants (MCP)
Talk to the RED Atlas API in natural language. Our remote MCP (Model Context Protocol) server lets AI assistants such as Claude query live market data, run valuations and read this documentation on your behalf, using your own account and plan.
Connector URL
Add this URL wherever your AI client asks for a remote MCP server:
https://apiv3.atlas.red/external-api/v1/mcpAuthentication & billing
When you connect, you sign in with your developer portal email and password (OAuth). No API key is exposed to the assistant. The connection is tied to your account and can be revoked at any time.
Assistant calls are metered exactly like regular API calls: they consume your plan's monthly credits, and the aggregate and premium tools cost more per call than a plain record read. Track everything in the Usage dashboard.
Connecting from your AI assistant
The connector follows the open MCP standard, so the same URL works across assistants. It is added once per workspace; each user then signs in with their own developer-portal account.
Claude (web, desktop, mobile)
- Open Settings > Connectors and find RED Atlas API in the connector directory — it is listed publicly, so there is no URL, client ID or secret to paste.
- Click Connect, sign in with your developer-portal email and password, then click Authorize.
- In a chat, open the connector menu to allow the tools you want — the read-only query tools can be left on "always allow"; report creation asks each time.
Adding it by URL instead (a self-hosted client, or any assistant that does not read Claude's directory): choose Add custom connector, paste the connector URL above and leave OAuth Client ID and Secret empty.
On Claude Team/Enterprise plans only an organization Owner can add the connector (Organization settings > Connectors). Keep Individual sign-in enabled so every member connects with their own RED Atlas account.
Claude Code
One command registers the connector by URL — the directory listing is not used by the CLI. Authentication opens in your browser:
claude mcp add --transport http redatlas https://apiv3.atlas.red/external-api/v1/mcp
# then, inside a session: /mcp -> redatlas -> AuthenticateChatGPT (OpenAI)
- In ChatGPT, open Settings > Connectors (available on Plus, Pro, Business and Enterprise) and choose Add custom connector / MCP server.
- Paste the connector URL; leave any client id/secret blank.
- Authorize with your developer-portal account, then enable the connector in a chat or project.
Gemini
Gemini CLI connects over MCP. Add the server to your ~/.gemini/settings.json (or a project .gemini/settings.json); the OAuth sign-in opens in your browser on first use.
{
"mcpServers": {
"redatlas": {
"httpUrl": "https://apiv3.atlas.red/external-api/v1/mcp",
"oauth": { "enabled": true }
}
}
}Grok & X
- In Grok (grok.com) or X, open Settings > Connectors > New connector > Custom.
- Paste the connector URL as the remote MCP server.
- Complete the OAuth sign-in with your developer-portal account.
Smithery
Smithery is a public MCP directory whose gateway proxies your client to this server, so it works from any Smithery-compatible client without pasting a URL. The listing is RED Atlas API on Smithery.
- Open the listing and click Connect — there is nothing to configure, because access is negotiated over OAuth rather than an API key.
- Sign in with your developer-portal email and password and click Authorize. Calls are metered against the same plan credits as every other client.
Available tools
The connector exposes the documented API surface as 18 tools. The assistant picks them automatically based on your question. The tool list is discovered over MCP at connection time, so tools we add show up without any change on your side — reconnect the connector if an assistant is still on an older list.
| Tool | Description |
|---|---|
| atlas_query_data | List records (properties, listings, parcels, ...) with filters, sorting and pagination. Supports the near radius and bbox viewport filters. Consumes credits per record. |
| atlas_get_record | Fetch a single record by id. Consumes credits. |
| atlas_get_stats | Aggregated market statistics, optionally scoped to a radius (near) or a map viewport (bbox). Fixed credit cost (5-15 depending on the entity). |
| atlas_market_report | Profile a market in one call: stats, a 12-month trend and an optional ranked breakdown. Fixed credit cost. |
| atlas_top_by | Rank and count records by a dimension, e.g. listings per municipality or per brokerage. Fixed credit cost. |
| atlas_pivot | Cross-tabulate two dimensions; either axis may be a date, which returns a time series per group. Fixed credit cost. |
| atlas_velocity | How fast a market is moving: new records this period vs the previous one, with the absolute and percent change. Fixed credit cost. |
| atlas_search | Free-text search across entities. Consumes credits per call. |
| atlas_lookup_location | Resolve city, neighborhood and region names into filter ids. Free. |
| atlas_list_entities | Discover which entity types a market offers. Free. |
| atlas_get_parcel_contacts | Parcel contact enrichment. Flat 3 credits per call, regardless of how many contacts come back. |
| atlas_enrich_contacts | Contact enrichment (skip-trace, premium) — owner/occupant contact data for a person or address. Charges the contacts_enriched credit price per call when contacts are delivered; no-match costs 0. Repeat lookups are served from the RED Atlas store. |
| atlas_create_avm_report | Start an automated valuation (AVM) report. 10 credits for JSON, 15 with the PDF. |
| atlas_create_comparables_report | Start a market comparables report. 6 credits. |
| atlas_comparable_sales | Comparable recorded sales near a property, with distance, price per m2 and the medians already computed. Runs the comparables report and waits for the result, so it is one call instead of create-then-poll. 6 credits. |
| atlas_asset_class_report | Profile one asset class in a single call: for-sale and for-rent aggregates, a building-class split, top municipalities, velocity and the total stock of that use. Billed as the ~7 aggregator calls it runs. |
| atlas_list_reports | List your generated reports. Free. |
| atlas_get_report | Fetch a report result and its shareable link. Free. |
| atlas_read_docs | Read this documentation so answers cite real parameters. Free. |
Things you can ask
- How many houses are for sale in San Juan under $500,000?
- Run an AVM for this address and give me the estimated value.
- Write a Python script that pulls last month's transactions in Condado.
Other AI clients
The connector follows the open MCP standard and also works with other MCP-capable clients, such as ChatGPT custom connectors and MCP-enabled IDEs. Point the client at the same URL and sign in with your developer portal account.