Skip to main content
Point any MCP-capable client at the AskFutures MCP server and you can build and backtest futures strategies right from your editor or terminal. The remote endpoint is:
URL
https://mcp.askfutures.com/mcp
You need MCP access enabled on your account first. There’s no self-serve toggle yet — see Authentication to request it. Without it, the connection will sign in but tool calls won’t be authorized.

​
Connect your client

1

Add the server

Use the snippet for your client below. They all point at the same remote URL over streamable HTTP.
claude mcp add --scope user --transport http askfutures https://mcp.askfutures.com/mcp
Add Cursor’s entry to its MCP config (Settings → MCP, or mcp.json), and Codex’s to its MCP servers config. Exact file locations vary by client version — check that client’s MCP docs.
2

Sign in

The first time the client connects, it opens a hosted AskFutures sign-in page in your browser. Sign in there and approve access. Your client stores the resulting token and rotates it automatically — you won’t be asked again until it expires. See Authentication for the details.
3

Verify the tools are available

Ask your agent to list the AskFutures tools, or check your client’s MCP panel. You should see tools for creating strategies, running backtests, and optimizing. If you only see them after signing in but calls fail, your account likely needs MCP access granted.
4

Run something

Try a plain-English request:
“Using AskFutures, build a strategy that buys Micro Nasdaq when price breaks above the first 15-minute high, then backtest it on the last year.”
Your agent will create the strategy, wait for the backtest to finish, and report the results.

​
Local option (advanced)

The MCP server can also run locally over stdio instead of connecting to the remote URL. This is mainly for development and self-hosting — most people should use the remote endpoint above, which is managed and always up to date. The remote connection is the supported path and what the rest of these docs assume.

​
What to do next

Once you’re connected, the workflow is create → wait → optimize: tools that do heavy work return a task you poll until it’s done. The overview explains the async model, and the recipes show full end-to-end examples.
Backtests are hypothetical and simulated, net of modeled slippage and commission. Past performance does not guarantee future results. Always test before you trade.

​
Next steps

Authentication

Sign-in flow, token lifetimes, and getting access enabled.

MCP tools

Every tool, what it does, and what it returns.

Recipes

End-to-end examples: build, backtest, optimize, compare.