If you're building an autonomous trading or betting agent, you've probably hit this friction: your data source is a dashboard, but your agent lives in a chat loop. You end up writing glue code to bridge the two.
I just shipped Signal Hub MCP, a small Apify Actor that closes that gap for two data sources I already maintain: unusual options activity (volume/open-interest anomalies on US equities) and horse racing odds-movement signals.
What it does
It's a standard MCP server, deployed as an Apify Actor in Standby mode (Streamable HTTP, not stdio). Two tools exposed:
- get_unusual_options_activity(tickers, min_vol_oi_ratio, max_results)
- get_horse_racing_signals(max_races, only_signals)
Your agent just calls the tool mid-conversation. No dashboard, no manual actor runs, no dataset parsing.
Why Standby mode + Pay-Per-Event
Apify's Actor Standby mode keeps the server warm and reachable over HTTP instead of spinning up per-request. Billing is Pay-Per-Event, charged per tool call, not a subscription. That mapped cleanly onto the MCP tool-call model: one Actor.charge() per tool invocation.
Internally, each tool call triggers a run of one of my existing standalone actors (Unusual Options Activity Scanner, Horse Racing Pulse) and returns the dataset. No duplicated scraping logic — the MCP layer is purely a thinner, agent-friendly interface on top of what already existed.
Try it
https://apify.com/0xgollum/signal-hub-mcp
More tools from the same portfolio (crypto funding rates, prediction markets) are next in line for the same treatment.
Top comments (0)