Skip to content

URSA: Universal Research and Scientific Agent

URSA is a flexible agentic workflow for accelerating scientific tasks. It helps connect language models, research tools, code execution, planning, persistent agent state, retrieval-augmented generation (RAG), and Model Context Protocol (MCP) servers into reusable scientific workflows.

Use URSA when you want to:

  • chat with a research assistant in the terminal or dashboard,
  • ask an agent to plan a technical task,
  • ask an execution agent to write, edit, and run code in a workspace,
  • connect to OpenAI, Anthropic, Google GenAI, Ollama, or an OpenAI-compatible endpoint,
  • persist named agents and reuse them across sessions,
  • ingest documents into persistent RAG collections,
  • compose multiple agents into teams or symposium-style peer review environments.
  • expose URSA as an MCP server for another client or agent framework,

Execution and network access

Some URSA agents can write files, run shell commands, and use web or MCP tools. Use a dedicated workspace, review actions carefully, and read the Sandboxing and information-control guidance before running high-trust or data-sensitive workflows.

Quick install

We recommend installing with uv:

uv tool install --python 3.13 'ursa-ai[dashboard]'

See Getting started for installation alternatives and a walkthough of using URSA.

Quick first run

For OpenAI, set the standard API-key environment variable and run URSA. The built-in provider includes the model and base URL, so no config file is needed:

export OPENAI_API_KEY="..."
ursa

$env:OPENAI_API_KEY = "..."
ursa

Inside the URSA app, type / to browse commands or try:

  • Summarize what URSA can help me do.
  • #execute Write and run a Python script that prints the first 10 prime numbers.

Where to go next