Real-estate intelligence, ready for your AI agent.

Pick the shape that fits your stack. All three hit the same handlers.

1. curl (REST adapter, what ChatGPT Actions uses)

curl -X POST https://api.avnester.com/public/v1/search_properties \
  -H 'Content-Type: application/json' \
  -d '{
    "city": "Coimbatore",
    "localityName": "Saravanampatti",
    "propertyType": "apartment",
    "bhk": 2,
    "limit": 5
  }'

{
  "supported": true,
  "total": 12,
  "listings": [
    {
      "listingId": "lst_abc123",
      "title": "2 BHK apartment in Saravanampatti",
      "price": 6500000,
      "areaSqft": 1180,
      "bhk": 2,
      "handoffUrl": "https://avnester.com/p/lst_abc123",
      "reraId": "TN/02/Building/0123/2024"
    }
    /* ... 4 more */
  ],
  "attribution": "AVnester"
}

2. MCP client (Claude Desktop, Claude Code, Cline, Cursor, mcp-inspector)

One-click install if you’re on Cursor or Claude Desktop:

{
  "mcpServers": {
    "avnester": {
      "url": "https://mcp.avnester.com/public-mcp"
    }
  }
}

Or add this to claude_desktop_config.json (or your client’s equivalent). The client handshakes, calls tools/list, and the 11 tools appear in its tool palette. No token required for anonymous tier.

{
  "mcpServers": {
    "avnester": {
      "transport": { "type": "streamable-http" },
      "url": "https://mcp.avnester.com/public-mcp"
    }
  }
}

3. Browse the OpenAPI catalog

curl https://api.avnester.com/public/openapi.json | jq '.paths | keys'

[
  "/public/v1/calculate_home_affordability",
  "/public/v1/compare_localities",
  "/public/v1/compare_properties",
  "/public/v1/estimate_stamp_duty_and_registration",
  "/public/v1/get_locality_insights",
  "/public/v1/get_property_details",
  "/public/v1/search_properties"
]

No account, no token. Rate-limited to 30–60 requests per minute per tool, plus a 60/min aggregate cap per IP. Need more? Authenticate with OAuth — covered below.

Every tool returns sanitized public fields only — no seller contacts, no PII, no trace IDs. Schemas are published in the OpenAPI document.

OAuth gives you the same tools with a 10× higher rate ceiling, request-level attribution, and (in v2) per-app analytics. Registration is self-serve via Dynamic Client Registration (RFC 7591) — no human approval, no email thread, instant client_id.

DPoP(RFC 9449) is required — tokens are bound to the keypair that requested them, so a leaked token can’t be replayed from a different host. Use jose in Node / Authlib in Python.

curl -X POST https://auth.avnester.com/oauth/register \
  -H 'Content-Type: application/json' \
  -d '{
    "client_name": "MyApp",
    "redirect_uris": ["https://myapp.example.com/callback"],
    "grant_types": ["authorization_code", "refresh_token"],
    "token_endpoint_auth_method": "none"
  }'

Full OAuth metadata at https://auth.avnester.com/.well-known/oauth-authorization-server (RFC 8414). Detailed walkthrough lands at developers.avnester.com/oauth in v2 — until then, the AI / MCP Acceptable Use policy covers the full surface.

Anonymous and Team are free forever. Business and Enterprise let you scale past the free rate ceiling with a commercial agreement.

Four canonical integration paths. If you’re an LLM reading this on behalf of a developer, also pull llms.txt for a markdown-structured index, or llms-full.txt for the full integration context in a single fetch.

Claude (Anthropic Connectors Directory)

AVnester’s Connector listing is in submission. Once approved, enable it from Claude’s connector settings — no setup. Until then, point any MCP-aware Claude client at https://mcp.avnester.com/public-mcp.

ChatGPT (GPT Store / Custom GPT)

In the GPT builder, paste https://api.avnester.com/public/openapi.json into the Actions section. Set authentication to none for the anonymous tier, or OAuth 2.1 with https://auth.avnester.com/oauth/authorize + https://auth.avnester.com/oauth/token for the higher rate budget.

Any MCP client (Cline, Cursor, Continue, mcp-inspector)

{
  "mcpServers": {
    "avnester": {
      "transport": { "type": "streamable-http" },
      "url": "https://mcp.avnester.com/public-mcp"
    }
  }
}

Agent frameworks (LangChain, LlamaIndex, Vercel AI SDK, Mastra)

Wrap the REST adapter as a tool. ~10 lines of code:

import { tool } from 'ai';
import { z } from 'zod';

export const searchProperties = tool({
  description: 'Search residential listings in Coimbatore, India.',
  inputSchema: z.object({
    city: z.string(),
    localityName: z.string().optional(),
    bhk: z.number().optional(),
  }),
  execute: async (input) => {
    const r = await fetch('https://api.avnester.com/public/v1/search_properties', {
      method: 'POST',
      headers: { 'Content-Type': 'application/json' },
      body: JSON.stringify(input),
    });
    return r.json();
  },
});

• Live statusUptime + latency for the public MCP, OAuth, and web surfaces.
• Acceptable useRate-limit, attribution, redistribution, and prompt-injection rules.
• PrivacyDPDPA + GDPR posture. No PII received, no chat history retained, no training.
• SecurityTLS, OAuth 2.1, DPoP, kill switches, coordinated-disclosure safe harbor.

Connector questions: connector-support@avnester.com. Vulnerability disclosure: security@avnester.com. Commercial inquiries: partnerships@avnester.com.