Local cache and auto sync
Market, account, order, and position data are cached locally and kept up to date automatically, so common actions respond fast.
Polytrader2Fast local desktop trading app. Open source custody. MCP-powered AI agents for strategy and bot automation.
Features
Polytrader2 keeps data, keys, and strategy execution local, while connecting AI agents through MCP.
Market, account, order, and position data are cached locally and kept up to date automatically, so common actions respond fast.
Keys are encrypted by the operating system, using DPAPI on Windows and Keychain on macOS.
Strategies run locally inside an isolated VM with controlled access to app capabilities and trade actions.
Connect Claude Code, Codex, and other AI agents to market context and strategy workflows through the built-in MCP Server.
Demo
From fast market browsing and local wallet records to AI-assisted strategy writing and bot execution, this section is ready for the product demo video.
MCP
Polytrader2 exposes a local HTTP MCP server. The default endpoint is http://localhost:8708/mcp and requests are protected with a Bearer token.
Run this command in your terminal to add Polytrader2 as a user-scoped MCP server.
claude mcp add --transport http --scope user polytrader2 http://localhost:8708/mcp \ --header "Authorization: Bearer <YOUR_MCP_TOKEN>"
Strategy
Polytrader2 is designed for real trading workflows: use local AI agents to help draft strategies, bind them to markets and wallets, then let the desktop app run, record, and contain execution.
Strategies receive account summaries and host-managed trading capability, while key material remains under local desktop custody.
Strategies run in an isolated VM child process and cannot freely access the desktop app, database, or system environment.
Orders and cancellations go through the desktop trading service and are written to order records.
The execution boundary, account handling, and order recording logic can be inspected in the source code.
Strategies use locally synced markets, order books, trades, orders, and positions, reducing repeated waits on remote APIs.
MCP gives Claude Code or Codex market context, so local AI agents can help draft, revise, and explain trading strategies.
A strategy can be bound to a specific market, outcome, and account, then run as a bot that reacts to data changes.
Starts, stops, logs, errors, and order requests are recorded locally for review and debugging.