Integration options

For a new integration, the main path is the server-to-server API: your server talks to Pereval, and the user interface stays in your product. An OpenAPI client and an AI agent help implement that same API; they are development tools, not separate ways to connect your users. Ready request flows by product type are here: swipe app, search site, bot.

Option	For	Status
Main approach: server-to-server API + events	Apps, websites and bots with their own backend	Available
Development tool: OpenAPI-generated client	A typed client for calling the same API	Available
Development tool: AI agent	Help writing an integration from the docs	Available
AI-agent tool: MCP (`POST /mcp`)	The agent calls the network with an API key; not a replacement for user-facing integration	Available

Server-to-server API

The main and most flexible option. Your server calls the API and receives events — through webhooks or with GET /v1/events; the app talks only to your server. You fully control the interface: how network profiles look, where the feed appears, how notifications read.

What you’ll need:

Store the network_id of your users who are in the network (it is returned in the PUT response).
A table of conversations with people from the network: match_id → your conversation.
A webhook handler with signature verification and duplicate protection. Without a webhook URL — regular pulling of events with GET /v1/events.
In the interface: consent toggles, a “from app X” label, an “AI-translated” label, a warning on suspicious messages.

First, test the API in the sandbox; then build the user flow using the guide for your product type.

OpenAPI-generated client

The specification is at /openapi.yaml (OpenAPI 3.1, including webhooks). You can generate a client from it in almost any language:

npx @openapitools/openapi-generator-cli generate \
  -i https://pereval.dating/openapi.yaml \
  -g php -o ./pereval-client

Instead of php, use python, typescript-fetch, go, java, kotlin, swift5, csharp and other generators.

AI agent working from the docs

If an AI agent (Claude Code, Cursor, Copilot and similar) writes the integration, give it:

/llms-full.txt — all documentation and the specification as a single text;
/openapi.yaml — the exact description of requests and events.

A ready-made task prompt for the agent is on the For AI agents page. Every documentation page has a “Copy for AI agent” button.

MCP — available

The same partner key as REST, POST /mcp. JSON-RPC 2.0, protocol version 2024-11-05. Methods: initialize, tools/list, tools/call, ping. Tools: upsert_member, get_feed, like, dislike, send_message, list_events, list_languages. A tool result is JSON in result.content[].text. Do not put the key in client-side code.

Need help

Use the contact form — we’ll help you choose a path, review your data model and answer integration questions. Read replies and continue the conversation on this site using the private link you must save after submitting. Incoming email is not accepted.