Vibeleaderboard
Index / tool
Visit github.com
Category
Developer Tools
Rank
No. 2046Tools index

Previous survey · No. 1983 ·

Pricing
Open Source
Type
TOOL
Use case
Coding · Software Testing & Quality
GitHub
4 stars
Latest release
v0.28.0
Date

About

A programming language and verification pipeline for experiments where AI agents write code under formal contracts. Agents fill typed holes, and an SMT solver proves each body before it merges, so incorrect fills are rejected rather than reviewed by hand. It is written with Haskell and Z3 and uses refinement types.

At a glance

LLMLL is an experimental language for having agents fill in functions that must satisfy formal contracts. It is worth exploring if you want to see how a verifier can reject code that typechecks but violates a requirement.

Good fit

Developers and researchers who can express a small invariant precisely and want to test contract-guided generation.

Consider something else when

Use your established language and test suite if you need a production ecosystem or cannot specify the key behavior as a contract.

Start here

Run the project’s balance-conservation example in a test environment. Compare the correct function with the one that creates an extra unit and inspect which postcondition rejects it.

Cost and requirements

The workflow requires the experimental LLMLL toolchain and its solver-backed verification path. Budget time for writing the contracts as well as implementing the functions.

Limits to know

A proof covers the contract supplied to the verifier. A missing or incorrect requirement can still leave a behavior unchecked.

Tags

programming-languageformal-verificationsmtai-agentshaskellrefinement-types

Tech Stack

Docker

Comments (0)

No comments yet

Editorially curated, with community endorsements as a secondary signal. Corrections welcome.