Assistants

When the Backtesting Agent has a project to work on, it works inside your project the way an experienced developer would.

The Backtest Assistant writes the algorithm, then compiles it and works through any warnings or errors the compiler surfaces before moving on. Once the code is clean, it runs a backtest and reviews the result for runtime errors, returning to the code to fix the root cause rather than papering over the failure. The cycle of write, compile, fix, run, and review repeats until the strategy executes end to end against historical data.

The code itself follows the conventions experienced QuantConnect developers use — modern universe selection, scheduled rebalancing, and the patterns that keep an algorithm readable and maintainable rather than fragile. The result is an implementation a quant could pick up, understand, and extend without first having to clean up after the author.

A working algorithm in your project, compiled clean, with a completed backtest you can open and read on your own terms. The Backtest Assistant will not summarize the results for you, because the backtest report already tells that story better than prose can, and it will never call the code production-ready — the next stop is paper trading, where the strategy gets to prove itself against live market behavior rather than history alone.

The Backtest Assistant has access to the following tools:

• create_compile
• read_open_project
• update_project
• read_project_nodes
• create_backtest
• read_backtest
• list_backtest
• update_backtest
• delete_backtest
• search_backtest_logs
• create_optimization
• read_optimization
• list_optimizations
• abort_optimizations
• delete_optimizations
• datasets
• user_input
• environment_library_support
• select_kanban_card
• object_store_get
• object_store_set