Skip to main content
The openclaw adapter wakes an OpenClaw agent over HTTP. The agent runs remotely and communicates back to AoA via its OpenClaw endpoint.

When to Use

  • Your agent runs as an OpenClaw server (remotely or on another machine)
  • You want streaming execution via SSE in a single AoA run
  • You want wake-style webhook callbacks

When Not to Use

  • You need local CLI execution on the same machine — use claude_local, codex_local, cursor, or opencode_local
  • The OpenClaw endpoint is not reachable from the AoA server

Configuration Fields

Core

Session Routing

Hire-Approved Callback

Called when this agent is approved/hired via the approval workflow.

Transport Modes

sse (default): AoA opens a long-lived SSE connection to the OpenClaw endpoint. The run stays open until the stream closes. Full stdout/log capture. webhook: AoA sends a wake payload and returns immediately. The OpenClaw agent processes asynchronously and calls back to AoA’s MCP endpoint. Use for fire-and-forget invocations or when the execution takes too long for a synchronous SSE session.

Wake Payload

OpenClaw receives a JSON payload merged from payloadTemplate plus AoA context fields: AOA_AGENT_ID, AOA_COMPANY_ID, AOA_API_URL, and AOA_RUN_ID. AOA_API_KEY is not injected — the OpenClaw adapter has supportsLocalAgentJwt: false. The remote agent authenticates back to AoA using its own configured API key.

Environment Test

The environment test checks that the OpenClaw URL is reachable and returns an expected response format.