NEWRead our new paper

Welcome to
CodeLogician™

Agentic governance framework bringing logical rigour and scalability to LLM-based software development.

Step 1:

Your first step is to obtain Imandra Universe API key that your instance of CodeLogician will use for connecting.

Get API Key

No credit card required.

Step 2:

pip install codelogician
# codelogician --help

Powered by:

Announcement

New paper is out!

Read the Paper

CodeLogician Performance Metrics

Comprehensive evaluation of LLM performance across multiple models and metrics

Model Comparison Across Metrics

Comprehensive radar view of all models' performance across seven evaluation metrics

Neurosymbolic AI for Code

For those who love AI coding, but still want to be in control. CodeLogician™ (CL) helps your AI coding assistant stay grounded in logic, explain its actions and scale beyond training data.

Formal Verification

Mathematically prove the correctness of your algorithms and systems, ensuring they meet specifications with absolute certainty through rigorous formal methods.

Test Case Generation

Automatically generate comprehensive test suites that cover edge cases and critical scenarios, ensuring thorough validation of your code and systems.

Logical Reasoning Booster

Empower your coding LLMs with logical reasoning so they can escape the limits of the training data and answer questions about your code based on logic, not the patters they have seen before.

CodeLogician™ helps coding assistants think logically

CodeLogician™ applies neurosymbolic AI to translate source code into precise mathematical logic, striving to create a formal model of the program's behavior that's functionally equivalent to the original source code. This model can then be analyzed with its reasoning tools (reasoners and agents) to prove deep properties, uncover hidden bugs, and automatically generate rigorous test cases.

Which reasoner or agent is best for reasoning about your code? It depends! At Imandra, we've created ImandraX - automated reasoning engine which we successfully applied to some of the world's most complex software systems in highly regulated environments and government agencies. But there're many other reasoners and tools (e.g. TLA+ and Lean) that we're working on bringing to CodeLogician™. If you have a specific request - please reach out and we will prioritize accordingly!

How it works

Neurosymbolic AI at scale

CodeLogician integrates neurosymbolic AI to revolutionize software development, combining the power of machine learning with formal reasoning.
Neurosymbolic AI at scale

Autoformalization

The latest CodeLogician builds on its agent-only predecessor applying autoformalization techniques to a new scale. It analyzes your entire project and creates a structured logical representation of your project - called the MetaModel.
Autoformalization Process

MetaModel Build-out

MetaModels capture logical structure and relationships within your project, enabling systematic logical reasoning and analysis.
MetaModel Build-out

Select Features

Formal Verification

Mathematically prove the correctness of your algorithms and systems, ensuring they meet specifications with absolute certainty through rigorous formal methods.
Formal Verification

Test Case Generation

Automatically generate comprehensive test suites that cover edge cases and critical scenarios, ensuring thorough validation of your code and systems.
Test Case Generation

Change Planning
with Sketches API

Plan and verify system changes using our Sketches API, enabling safe evolution of complex systems through formal reasoning.
Change Planning with Sketches API

Underlying
technologies

Neurosymbolic AI
for Code

CodeLogician™ is powered by ImandraX, our award-winning automated reasoning engine that combines formal verification, symbolic reasoning, and rule synthesis to provide mathematical guarantees about your code's behavior.

Industrial automated reasoning engine

The latest version of our automated reasoning engine for algorithm analysis, bringing unprecedented rigor and automation to algorithm design and governance.
Read more

Real-world industrial applications in safety critical industries

ImandraX is widely used in safety-critical industries such as financial markets, robotics and autonomy, and model-based systems engineering in government and defense.
Read more

Formal verification

Mathematical verification of algorithm properties with powerful reasoning automation powered by ImandraX -- bringing formal proofs and counterexamples seamlessly into the software engineering workflow.
Read more

Region decomposition

Inspired by Cylindrical Algebraic Decomposition, this technique decomposes algorithm state-spaces into regions where system behavior is invariant, effectively computing a map of all possible system behaviors.
Read more

Seamless integration of bounded and unbounded verification

Seamlessly combine bounded model-checking and unbounded inductive proof through ImandraX. Every goal can be subjected to both bounded verification (complete for counterexamples in a precise sense) as well as unbounded verification, e.g., using Imandra's decision procedures and higher-order lifting of Boyer-Moore style automated induction.
Read more

Counter-examples

ImandraX is complete for counterexamples in a precise sense. Automatically synthesizes deep counterexamples to incorrect verification conjectures, even with recursive functions and nonlinear arithmetic, aiding discovery and eliminating time sinks from trying to prove false goals.
Read more

Imandra Discover

Discover program properties and lemmas, extract logical patterns from data (e.g., logs) and synthesize them into executable and auditable models. Identify anomalies in out-of-sample data and generate corrective actions for detected issues.
Read more

Easy Setup

1 minute to Mars start

Step 1

Your first step is to obtain Imandra Universe API key that your Instance of CodeLogician™ will use for connecting.
Get API key

Step 2

Install CodeLogician™ using pip.
> pip install codelogician

Step 3

Start using CodeLogician™ to analyze your codebase.
> codelogician --help
VSCode Extension