A workbench for mathematicians
and their agents.

Think through the mathematics. Let coding agents help build it. Glitnir brings Lean proofs, readable notebooks and computational exploration into one shared workspace.

Lean sourceGlitnir notebookYour coding agent
VS Code with Lean source on the left, the rendered Glitnir notebook in the middle, and a coding agent on the right
Read the reasoning as your agent works. Follow changes, inspect proofs and steer the next step.

Stay close to
the mathematics.

Coding agents can write Lean. Glitnir makes their work easier to read, question and develop together.

A notebook that is still a Lean file.

The Glitnir notebook specification adds Markdown prose and mathematical notation inside Lean comments. Definitions and proofs remain ordinary Lean, in the same file.

Your agent writes the source. You read the argument. Both work with the same document.

Read the notebook format
Hello.lean Minimal notebook example
/-! # A first notebook

Explain the idea, then give its proof.
-/

theorem hello : True := by
  trivial
Readable prose. Standard Lean. One source of truth.

Read and review in VS Code

Open a notebook beside its source. Follow live updates from your agents, jump between prose and code, explore proof steps and check saved claims independently with Lean.

Explore the extension

Bring your coding agent

Install the notebook authoring skill for Codex, Claude Code or another coding agent. Give it the notebook format and checking rules, then ask it to explain its work as it goes.

Get the agent skill

Make the mathematics visible

Bring graphical views, simulation outputs and other artifacts into your notebook. The optional Forseti plugin connects supported Lean models to visualizations and simulations.

Meet the Forseti plugin

Know what has been checked

Keep checked proofs, unfinished arguments and additional assumptions visible. Simulations help you explore a conjecture; Lean checking tells you what the proof establishes.

Understand claim checking

Coming next

A bigger team.
More room to compute.

We’re extending the workbench with Taskvalet: coordinated agent research and shared computational resources, connected to your notebooks.

Agent swarms with shared memory

Bring a swarm of agents to one mathematical problem, with persisted findings, shared context and coordinated work. Coding-agent skills and packaged tools for harness agents will connect their contributions to Glitnir notebooks.

Compute for proofs and simulations

Offload heavy work to distributed compute optimized for Lean verification and simulation. The goal: let people and agents submit demanding jobs while agents keep working on proof construction.

Public signup for swarm and compute orchestration is not open yet.Follow development on GitHub

Start with a notebook.

Install Glitnir, open a Lean file and bring your agent into the conversation.

Get Glitnir for VS Code

Setup needs uv and Python; checking needs compatible Lean.
See installation requirements