# lean-dojo/LeanCopilot

LLMs as Copilots for Theorem Proving in Lean

Repository: https://github.com/lean-dojo/LeanCopilot
Canonical: https://ross.abutalabs.com/products/leancopilot
Homepage: https://leandojo.org/leancopilot.html
Language: C++
License: MIT
License Family: permissive
Topics: lean, llm-inference, machine-learning, theorem-proving, lean4, formal-mathematics, llm
Last push: 2026-08-22T02:14:29+00:00

## Health v2 (maintenance only)
Score: 90/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 98, release rhythm 86, longevity 77
- inputs: {"age_days": 1090, "days_push": 12, "days_rel": 14, "gap_med": 32.5, "n_releases_24m": 17}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1316, forks 127 (observed 2026-08-28T04:04:20.855313+00:00)

## What it is
Lean Copilot integrates large language models natively into the Lean 4 theorem prover for proof automation. It provides tactic suggestion, premise selection, and proof search using built-in models from LeanDojo or user-supplied models running locally or in the cloud.

## Use cases
- suggest tactics for a Lean proof I'm stuck on
- search for complete proofs automatically in Lean 4
- find relevant theorems and lemmas to use as premises
- run an LLM copilot inside my Lean project
- bring my own model for theorem proving assistance
- accelerate formalizing mathematics in Lean

## When to choose
- you write proofs in Lean 4 and want AI-assisted tactic suggestions or proof search
- you want to plug custom local or cloud-hosted LLMs into Lean
- you work on formal mathematics and want premise retrieval from Mathlib-style libraries

## When to avoid
- you need a general-purpose code completion copilot outside of Lean
- your project targets Lean 3 or versions below lean4:v4.3.0-rc2
- you cannot install native dependencies like CTranslate2 or lack supported platform support

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, machine-learning, plugin-system, sdk
- domain: artificial-intelligence, large-language-models, mathematics, developer-tools, education
- platform: windows, cpp, python
- tags: lean4, theorem-proving, formal-mathematics, tactic-suggestion, proof-search, premise-selection, ctranslate2, human-ai-collaboration, linux, macos

## Member repositories
- lean-dojo/LeanCopilot (main) score 90

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:20.855313+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:48:29.837836+00:00, confidence not recorded.
  - readme: https://github.com/lean-dojo/LeanCopilot (fetched 2026-08-28T04:04:20.855313+00:00, sha 25ee4f4f1935)
  - homepage: https://leandojo.org/leancopilot.html (fetched 2026-08-29T12:06:55.383881+00:00, sha 607ef9b4ec3e)
- Data as of 2026-08-30T08:39:29.467469+00:00.