> ## Documentation Index
> Fetch the complete documentation index at: https://www.agentworldprotocol.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Wrap a simulator

> A lockstep adapter over an existing sim, with seeding, snapshot, and replay.

<Steps>
  <Step title="Map the surfaces">
    Sim state → observation channels; sim inputs → action schemas; sim step() → tick. Keep the mapping table in your repo — it becomes your manifest.
  </Step>

  <Step title="Implement lockstep semantics">
    Advance only on `world.tick` (AWP-TIM-002); stage submitted actions until the next advance (AWP-TIM-010); per advance, emit statuses and events first, then exactly one frame per per-tick channel, then the tick result (AWP-TIM-003/011); deliver a tick-0 frame right after `session.ready` (AWP-TIM-009); remove all wall-clock reads from the state path (AWP-TIM-004). If more than one lockstep session can bind, declare `tick_authority: "barrier"` (AWP-TIM-012).
  </Step>

  <Step title="Add sim-profile capabilities">
    Wire `seed` into your RNGs, implement `world.snapshot`/`restore` over your sim's serialization, and verify AWP-REP-003 by replaying a recorded log in CI.
  </Step>

  <Step title="Conformance">
    Map each requirement you implement to its ID in the [requirement matrix](/spec/requirements) and keep that mapping in your repo. The conformance suite that will automate this check is not published yet; until it is, a claim of conformance is self-assessed and MUST say so (AWP-CNF-005).
  </Step>
</Steps>

Planned reference implementations: [awp-sim](/adapters/awp-sim) and, for Gymnasium environments, a [bridge](/adapters/gymnasium-bridge) that does this mapping generically. Neither is published yet.