MAIA is a lightweight AI assistant designed to bring AI directly into the environments where you already work.
Primarily used from the command line, it works in a local terminal, over SSH, or in an IDE's integrated terminal. A GUI is also available, either started as a local web server or integrated into an existing web server.
MAIA provides AI models with structured access to sessions, workspaces, context and tools — while keeping the user in control of what the AI can see and what it can do.
Lightweight and portable
MAIA has no compiled components or large runtime environment. Built around standard command-line tools such as Bash, Perl, jq and curl, it can be deployed by copying the files to a system where the required tools are available.
This makes it suitable for a wide range of Unix-like systems, from development machines to remote servers, including Windows environments using WSL or Git Bash.
More than a chatbot
MAIA provides a structured environment in which an AI model can work with:
- Sessions — keep separate conversations and working environments isolated.
- Workspaces — give the assistant a defined working area.
- Context — explicitly control which files and other information are provided to the model.
- Tools — allow the assistant to interact with the environment through defined operations.
- Skills — extend the assistant with reusable capabilities and instructions.
- Instructions — project, tool and path matching instructions.
- MCP — connect MAIA to external tools and resources through the Model Context Protocol.
- Subsessions — allow separate MAIA sessions to cooperate when a task benefits from multiple independent contexts.
User-controlled context
MAIA does not assume that everything available on your computer should automatically become AI context.
You decide what information is made available to the model. Files, resources and other context can be explicitly selected and managed, while sessions provide isolation between different tasks.
This makes the context visible and controllable rather than relying on an opaque indexing system running in the background.
Controlled tool access
An AI assistant becomes significantly more useful when it can do things rather than only answer questions. It also becomes more important to understand what those actions can affect.
MAIA therefore treats tool capabilities according to their effects.
Operations such as modifying files or executing commands can require an inspectable change proposal before the operation is applied. Tool permissions determine which capabilities are available to the AI in the first place.
The goal is not to make the AI powerless. It is to make its capabilities explicit, inspectable and controllable.
Model and provider independent
MAIA is not tied to a particular AI provider or model.
The assistant can work with different AI APIs and providers, allowing you to choose the model and service appropriate for the task. The command-line environment, sessions, context and tools remain under MAIA's control rather than being tied to one provider's application.
Open source and auditable
MAIA is open source and released under the GPLv3 license.
The implementation is intentionally small and based largely on standard command-line tools. This makes the software itself inspectable rather than requiring you to trust a large, opaque runtime.
Get started
MAIA is available as source code and can be installed without compilation or a system-wide installer.
See the GitHub repository for installation instructions, configuration, documentation and the latest development information.