Connect an AI assistant
If you connect Claude, Codex, or another MCP-compatible assistant to your DevGhouse account, it can browse, trade, and manage your positions on Predictions directly, without you clicking through the site yourself. See Integrations: MCP for how to get a connection token; once connected, the assistant reaches Predictions the same way you do, scoped to your account.
What it can do
- Browsing, free for anyone:
prediction_list_markets(filter by category, status, search, or sort),prediction_get_market(one market's full detail and current price),prediction_get_price_history(24h, 7d, or 30d) - Creating a market:
prediction_create_market, debits your chosen initial liquidity in credits - Trading:
prediction_tradeto buy or sell shares on a side, optionally setting a take-profit and/or stop-loss in the same call - Managing an open position:
prediction_set_exitsto add or change a take-profit or stop-loss on a position you already hold - Your own results:
prediction_positions(open positions with live profit and loss),prediction_portfolio(total invested, total value, profit and loss, win rate) - Resolving a market you created:
prediction_resolve_market, paying out every open position, or refunding everyone if marked Invalid; only the creator can call this
For example: "Explain this market's settlement rule and my current exposure to it, do not place a trade" lets an assistant read and summarize without touching your credits.
Where this fits your workspace
Predictions tools sit in the platform's launchpad module, alongside the crypto trading tools. A run started without a specific project reaches your whole personal account.
Anything that spends credits, trades, creates a market, or resolves one asks for your explicit approval first, the same way a person would confirm before clicking a button. Discover the exact tools available to your connection with search_tools before relying on a specific one.
Want a standing agent instead of asking one-off questions? Deploy a trading agent wraps the same trading tools into a guided, capped, scheduled routine you set up once from the web app.