Developer Integration | Social Matchmaking API

Prepare behavioral features before matchmaking

Agree on the behavioral observations, game parameters, and voluntary repeat-play events your studio can provide and authorize for analysis. Historical or incremental exports build player and relationship features offline; recommendation requests use those features with live context rather than replaying full player histories.

The reference prototype uses deterministic features and rule-based scoring. Custom model training, evaluation, and reviewed model releases are planned as separate processes, not actions triggered automatically by a history upload.

Illustrative request workflow

Send eligible players

Provide eligible player references, live constraints, and current context to complement the behavioral features prepared offline.

Receive ranked groups

Request behavior-informed squad recommendations with factors your backend can inspect. Your game retains final formation and safety decisions.

Report outcomes

Report completed sessions and explicit voluntary regrouping. The planned learning loop also needs decision-to-outcome attribution before model training.

Example request

POST /v1/squads/recommend
{
  "game_mode": "ranked_squad",
  "region": "eu-west",
  "players": [
    {
      "player_ref": "opaque-player-reference",
      "role": "support",
      "play_style": "coordinated",
      "communication": "voice-preferred"
    }
  ]
}

Player references should be scoped and pseudonymous. Exact fields, retention, access, and deletion requirements must be agreed before any production integration.

Questions for the design-partner phase

Which formation decisions should the compatibility layer influence?
Which studio-supplied behavioral signals are available and authorized for analysis and training?
Which declared preferences and live context should complement those signals?
What latency and availability constraints must the recommendation meet?
How should studio safety policy affect eligibility?
How will voluntary regrouping be attributed to recommendations without counting accidental rematches?

Shape the first integration.

We are looking for multiplayer studios willing to define a concrete use case, integration boundary, and success measure.

Request developer access

Integrate social matchmaking without replacing your core queue.