AI assistant, knowledge search, and MCP server

Get help writing and fixing Python backtests with VBT documentation, code examples, and community answers close at hand. Ask a question in your notebook, search for a working example, or connect a coding assistant such as Claude Code, Codex, or Cursor to your VBT installation. Find the right API, test an idea, and inspect the results in the same conversation.

AI research assistant

Chat with an assistant grounded in VBT documentation and examples, with tools and reasoning

Knowledge search

Search documentation, API, examples, and community answers with keyword and embedding ranking

MCP server

Let your coding assistant search VBT documentation, look up the API, and run code over MCP

Choose where to start

What do you want to do?	Explore
Ask how to use VBT, fix a code snippet, or get help in a notebook	AI research assistant, with chat based on VBT documentation and examples, class-specific help, and callable tools.
Find an API, code example, or community answer and read the source yourself	Knowledge search, with keyword and natural-language search, source links, and support for your own research notes.
Work on a project with your existing coding assistant	MCP server, with documentation lookup, source inspection, agent skills, and Python execution in your installation.

From a focused question to a backtest

Start with the part you are working on, such as adding a trailing stop or selecting parameters on training data. ChatVBT retrieves relevant VBT material to help answer the question. You can include your code or an error message and continue with follow-up questions. Models that support tool calls can use vbt.interact to look up more information or run Python while answering. Supported models can also display reasoning summaries supplied by their provider.

For changes across project files, give your coding assistant VBT's documentation and agent skills. The MCP server lets it check the API, read examples, and run code. After a backtest, ask it to inspect the trades or compare settings while holding the rest of the strategy fixed. The MCP demonstration shows the search, API lookup, and execution behind one such task.

Use your choice of model

ChatVBT supports hosted providers and local models through Ollama. Search and chat providers can be chosen separately, and knowledge search can supply context for your own model. If you just need to find a warning or function name, keyword search works without a model provider once the knowledge assets have been downloaded.

Start here

Browse the Cookbook for practical recipes on loading, searching, and chatting with knowledge assets. The members-only Using AI guide covers agent skills, documentation access, and connecting your coding assistant.

features/ai pages