Next-gen copilot agent for your OS, powered by MCP
Stratapilot is a system-level, general-purpose copilot agent that automates and executes user tasks across applications, files, and interfaces. Designed for OS-native deployment, Stratapilot enables context-aware, multi-modal task execution through natural language input.
At the core of Stratapilot is the Model Context Protocol (MCP) — an extensible, application-agnostic protocol that allows Stratapilot to interface with third-party software (both legacy and modern) in a structured, deterministic way. MCP defines a standard for model-to-model and model-to-application communication, making Stratapilot interoperable across tools, services, and agents without requiring custom plugins.
Clone the repository:
git clone https://github.com/KAIST-KEAI/stratapilot.git
Set up Python environment (3.10+):
conda create -n stratapilot_env python=3.10 -y
conda activate stratapilot_env
Install dependencies:
cd stratapilot
pip install -e .
Configure your backend LLM API key:
Set your LLM provider's API key (e.g., OpenAI) in the .env
file and select your preferred model.
Run the demo:
python quick_start.py
Set up GUI mode (optional):
Specify the default agent prompt in a configuration file named prompt.conf
and place it in the root directory under the cloned project. You can find an example prompt file in prompt.py
. After this is done, enable GUI mode and start the program with
python quick_start.py --gui
The prompt in the configuration file is used to fine-tune the agent's behavior upon startup, and will not be visible in the chat history. You can then chat with the agent like with any LLM chat bot services.
Stratapilot provides a developer SDK for building:
Stratapilot supports a variety of integration methods out-of-the-box:
ffmpeg
, git
, curl
, etc.)See sample code under examples
for SDK usage demos and integration templates. For more information about tool development using the SDK, refer to Stratapilot's documentation.
Stratapilot welcomes contribution in any form! If you are interested in getting involved, refer to our contribution guidelines.
No configuration available
Related projects feature coming soon
Will recommend related projects based on sub-categories