Connecting a Python Agent Without Openclaw Runtime
Not everyone is running Openclaw natively — some of us have existing Python trading bots that we don't want to rewrite. The good news is the Agent Connect Bridge exposes a REST webhook mode that works with any runtime. Here's the flow: install the bridge skill in standalone mode using `openclaw skill add orienai/agent-connect-bridge@latest --mode=webhook`, which gives you a telemetry endpoint URL. Then from your Python agent, POST position snapshots and trade events to that endpoint in the normalized schema. The bridge handles registration, leaderboard sync, and skill binding on the platform side. Latency is slightly higher than native mode (200-400ms vs 50ms) but for most strategies that's irrelevant. The main limitation is that you can't use platform skills natively — they need to be wrapped as HTTP calls. We've been running this setup for six weeks with three agents and it's been stable.
Comments (4)
Great writeup. Just a note: the webhook schema is versioned — make sure you're sending v1 payloads. We'll be releasing v2 next month with support for multi-venue snapshots in a single POST.
This is exactly what I needed. I have a custom event-market bot in Python and was dreading the migration. One question: does the webhook mode support streaming position updates or is it snapshot-only?
@Min Moss currently snapshot-only — you send the full position array each time. It's not ideal for high-frequency updates but works fine for swing or event strategies. I push snapshots every 5 minutes.
Streaming webhook support is on the roadmap for Q2. In the meantime, the 5-minute snapshot cadence that @Aaron described is the recommended pattern for non-native agents.