Build your own endpoints that appear automatically in MCP Server — no server changes required
Custom endpoints created through the Custom Endpoint Builder are automatically available in the MCP Server — no server changes required.
The Anysite MCP Server dynamically loads all endpoints from the OpenAPI specification at startup. This means any custom endpoint you create through the Custom Endpoint Builder instantly becomes available to all your AI agents.
Use the [Custom Endpoint Builder](https://app.anysite.io) to define your parser — specify the target URL, data extraction rules, and output schema.
The MCP Server dynamically loads it from the OpenAPI specification. No server restart or configuration changes needed.
Your AI agent can find the new endpoint via `discover()` and call it via `execute()`:
```text
discover("your_source", "your_category")
→ Shows your custom endpoint with parameters
execute("your_source", "your_category", "your_endpoint", {params})
→ Fetches data using your custom parser
```
Custom endpoints appear instantly in MCP — no code, no deployment, no server updates.
Same caching, pagination (`get_page`), analysis (`query_cache`), and export (`export_data`) as built-in endpoints.
Works with Claude Desktop, Claude Code, Cursor, n8n, and any MCP-compatible client.
Custom endpoints are included in the MCP Unlimited Plan — no additional costs.
Suppose you've built a custom parser for a job board:
1. discover("my_jobboard", "search")
→ Shows: endpoint "search_jobs" with params: query, location, salary_min
2. execute("my_jobboard", "search", "search_jobs", {
query: "python developer",
location: "Berlin",
salary_min: 60000
})
→ Returns job listings + cache_key
3. query_cache(cache_key, sort_by="salary", sort_order="desc", limit=10)
→ Top 10 highest-paying positions
4. export_data(cache_key, "csv")
→ Download all results as CSV
Create your first custom endpoint at app.anysite.io