RACE AI is a research workspace with a built-in AI agent. You ask a question, and RACE searches the literature, reads papers, and writes an answer that links each claim to its source. From the same project, you can draft LaTeX or notes with an agent, run Python next to your data, and map a field with Knowledge Discovery. It's built for researchers in physics, biology, chemistry, and computing, plus labs that want to share projects. It runs in the browser or as a Mac and Windows desktop beta.

• Ask with Citations: Get answers drawn from OpenAlex, arXiv, Semantic Scholar, Crossref, and Europe PMC. Every claim is cited, and you can open each source in one click.
• Write Agent: An agent edits your LaTeX, notes, or code by proposing changes. You accept or reject each line, so you stay in control of the final text.
• Run Python in the Project: Python runs inside your project, right next to the papers and data it uses, so you don't need a separate environment for analysis.
• Knowledge Discovery: Map the key papers, state of the art, open gaps, methods, datasets, and people in a field. RACE keeps the map up to date as you work.
• Library: Store your papers in the same project as your questions, drafts, and code.
• Collaboration and Team Memory: Share projects with co-authors, assign roles, and keep a shared team memory for the whole group.
• Connect Your AI: Run every RACE feature on your own OpenAI, Anthropic (Claude), Google Gemini, DeepSeek, or OpenRouter API key. Work on your own key doesn't use RACE credits.
• Model Choice per Project: Auto mode uses GPT-5.6 Terra for everyday work and GPT-5.6 Sol for deep research and coding. You can also pick a model per project, and Gemini and Claude appear when enabled.

• Claims You Can Check: Each statement in an answer links to its source paper, so you can verify it in one click instead of trusting an unsourced summary.
• One Project for the Whole Workflow: Literature search, writing, Python, and field mapping all live in one place. You don't have to copy material between a chat tool, an editor, and a notebook.
• You Approve Every Edit: The writing agent proposes line-by-line changes for you to accept or reject, which keeps you in charge of your manuscript or code.
• Bring Your Own Key: Connecting your own API key from five supported providers lets you control model costs and keep working without using RACE credits.
• No Training on Your Research: RACE uses providers' business APIs that don't train on your content, and files are encrypted in transit and at rest.

• No Published Pricing: The site mentions paid plans with more capacity but doesn't list prices or plan details, so you can't compare costs before signing up.
• Desktop App Is an Early Beta: The Mac and Windows app (v0.1.1 beta) isn't code-signed, doesn't update automatically, and only supports email/password sign-in for now. Teams with strict IT policies may prefer the browser version.
• Race Lab Is Not Available Yet: The multi-agent feature for splitting open problems into subtasks and checking math with formal proofs is only a preview. It will start with theorems that are already proven.
• Department Licensing Still in Setup: Project sharing works today, but department licences are still being arranged through pilots. Institutions that need formal campus-wide contracts may have to wait.

• Literature Review for a New Project: Ask RACE a research question and get a cited answer drawn from major scholarly indexes. You can open each source directly and save the useful papers to your Library.
• Mapping an Unfamiliar Field: Use Knowledge Discovery to see the key papers, open gaps, methods, datasets, and active researchers in an area. It gives a new PhD student or a researcher changing fields a clear starting map.
• Drafting and Revising a LaTeX Paper: Let the write agent suggest edits to your manuscript or notes, then accept or reject each line. Your draft stays next to the sources it cites.
• Reproducible Analysis Next to the Papers: Run Python inside the project to analyze your data or reproduce a result from a paper. Code, data, and references stay together in one place.
• Lab Collaboration: Share a project with co-authors, assign roles, and build shared team memory. Everyone works from the same papers, drafts, and context.
• Cost-Controlled Research on Your Own Key: Connect an OpenAI, Anthropic, Gemini, DeepSeek, or OpenRouter key so your provider bills you directly and your work doesn't use RACE credits.