Update: An onchain market intelligence layer (Bitquery + OpenClaw)
This short update pulls together a thread on running an AI agent as a continuous, real-time market intelligence layer across onchain venues. The setup compresses latency and coverage by running parallel streams, applying consistent rules, and routing decisions through human approval while reducing manual tab-switching around the clock.
The compounding problem
Onchain traders hit two constraints at once:
-
Latency during active hours: Bots react in milliseconds. Humans monitoring and context-switching move slower, so edges during peak flow are easy to miss.
-
Availability across 24 hours: Polymarket, meme venues, and cross-chain arb rarely line up with sleep or a single screen. Off-hours, signals keep printing and often go unseen when nobody is at the keyboard.
What the layer is doing
-
Streaming in parallel: Keeping one clock across prediction markets, high-velocity meme ticks, and wrapped BTC on multiple networks so detection is not waiting on a human serial scan.
-
Applying consistent rules: In experiments: large Polymarket prints above a notional threshold trigger mirroring at half size; PumpFun tick jumps above X% trigger queueing a small exploratory buy. Criteria stay repeatable and written down.
-
Queueing then approving: The agent batches what crossed the line: “here are N signals that hit your rules; approve or reject.” The trader spends more time gating and refining and less time rebuilding the tape by hand.
Illustrative captures from that setup (terminal UI and logs; simulation-style IDs and books, for illustration only):
Stack (how the pieces map)
| Layer | Role |
|---|---|
| Bitquery | Supplies real-time structured streams: Polymarket fills, PumpFun USD OHLC, multi-network BTC, Trading.Tokens-style charts. The agent consumes APIs, not scraped HTML. |
| OpenClaw skills | Detection layer: small behaviors that load when the user asks for X and run subscription Y (BTC feed, multi-token USD chart, Polymarket trades, PumpFun, stablecoin flows). |
| OpenClaw agent | Decision and orchestration: keeps sessions alive, applies your rules, batches signals, and drives approve or reject (including the simulated books in the screenshots above). |
Where to install and read the code
All Bitquery-backed OpenClaw skills referenced in this note are open source and documented here:
github.com/bitquery/openclaw-skills-master
That repo documents installing via ClawHub (clawhub install …), setting BITQUERY_API_KEY, and running the Python streaming scripts for bitcoin-price-feed, crypto-chart-usd, pumpfun-usd-price-stream, polymarket-real-time-trades, and stablecoin-payments. For registry context, see ClawHub and OpenClaw.
On ClawHub I publish the onchain data streams I actually run—packaged as installable skills with the fields and filters:
- Reliable Bitcoin price feed (
bitquery-crypto-price-stream) - Crypto charting with USD pricing (1s) (
crypto-chart-usd) - Polymarket real-time trades stream (
polymarket-real-time-trades) - PumpFun USD price stream (
pumpfun-usd-price-stream) - Solana stablecoin payments tracking (USDC / USDT;
stablecoin-paymentsin the skills repo)