AI Writing Collaborator
Draft lecture notes, slide decks, papers, and private research notes with an AI that understands mathematical structure. It works alongside you in your document, not as a separate tool you have to context-switch to.
AI-powered mathematical research workspace
Write math. Run code. Deploy agents that research for you.
MathScribe is the AI-powered platform where mathematicians write, compute, and launch autonomous research agents — all inside a single shared workspace.
One platform for writing, computing, and autonomous research — no stitching tools together
Agents that keep running while you sleep, grounded in your own workspace context
Secure sandboxed code execution with every document
Document graph
theorem → code → figure → agent run
AI pass
Notes, computations, references, and agent runs stay attached to the workspace.
Inside the workspace
One workspace for writing, computing, and autonomous research
Live Code Execution
Run Python for symbolic algebra, numerical computation, and data analysis without leaving your document. Every execution happens in a secure, sandboxed environment so your workspace stays clean and your results stay reproducible.
2D Figures and 3D Models
Ask for a plot or a 3D model file and get one generated on the spot. Visualization is built into the writing flow, not bolted on after the fact.
Web Search, In Context
The AI assistant can search the web mid-session to pull in definitions, results, or references without breaking your focus. Everything it finds feeds directly into your document.
Agentic Research Harnesses
Design, test, and deploy multi-agent AI pipelines that run for hours or days on hard mathematics research tasks. You set the objective; the agents work through it autonomously, with human steering available whenever you need it.
Shared Agent-Human Workspace
Your AI agents have full access to your workspace, so every run builds on your existing notes, computations, and context. No more copying outputs between tools — the memory is shared from the start.
Workflows
One workspace for the full research loop.
MathScribe treats prose, code, figures, references, and agent runs as connected work instead of separate chores.
Write
Serious mathematical documents
Draft lecture notes, slide decks, papers, and private research notebooks without separating exposition from notation.
Compute
Python results beside the argument
Run symbolic algebra, numerical computation, and data analysis in a sandbox tied directly to the document.
Research
Agent runs grounded in workspace memory
Design, test, and deploy multi-agent AI pipelines that can keep working for hours or days with human steering available.
Secure Python
Run code where the argument is written.
Run Python for symbolic algebra, numerical computation, and data analysis without leaving your document.
Agent harnesses
Let hard research tasks keep moving.
Design, test, and deploy multi-agent AI pipelines that run for hours or days on hard mathematics research tasks.
Why MathScribe exists
MathScribe is built by Blue Note Lab for the people who do serious mathematics work: researchers pushing into open problems, instructors producing rigorous course materials, and anyone who thinks carefully about math and needs their tools to keep up. We built it because the existing stack — scattered LaTeX editors, disconnected CAS tools, general-purpose AI chatbots — asks you to manage the integration yourself. MathScribe puts the AI assistant, the Python kernel, the visualization engine, and the agentic research layer into one workspace with shared memory. The result is a research environment where human and AI contributors are genuinely working together, not trading files.
Questions mathematicians ask first
Practical answers, no generic SaaS gloss.
Who is MathScribe built for?+
Mathematicians at all levels who produce serious written work: undergraduate instructors building course materials, graduate students and researchers writing papers, and anyone running computational or theoretical investigations. If your work involves both writing mathematics and thinking hard about it, MathScribe is for you.
What kinds of documents can I create?+
Lecture notes, slide decks for undergraduate courses, research papers, and private research notebooks. The editor handles mathematical content natively, so you spend your time on mathematics rather than formatting.
How does the sandboxed code execution work?+
Python runs in a secure, isolated environment tied to your document. You write or ask for code, it executes, and the results appear inline. Nothing runs on your local machine and each execution is contained so one session can't affect another.
What can the research agents actually do autonomously?+
They can work through complex mathematics research tasks over multi-hour or multi-day runs — searching, computing, drafting, and reasoning — using your workspace as shared memory. You can steer them at any point or let them run to completion and review the output.
Is my workspace private?+
Yes. Your notes, computations, and agent runs are your own. Agents access your workspace to collaborate with you, not to share your work externally.
Do I need to know how to build AI pipelines to use the agentic features?+
No. MathScribe gives you a design and testing interface for building agent harnesses without requiring you to write infrastructure code. You define the task and the configuration; the platform handles the rest.
Private beta
Start with the math you are writing now
Tell us whether you are writing course material, a paper, a computational notebook, or a research-agent workflow. We will route you to the right MathScribe path.