Chelis Lang
Chelis is an open-source, statically typed functional language for numerical code written by AI coding agents and supervised by people. It builds tensor shapes, precision, effects, ownership, reproducibility, and property checking into the language/compiler so errors can be caught before execution or review.
At a Glance
- AI coding-agent and developer-tooling teams
- Researchers and engineers doing numerical computing
- Scientific-computing users
- Quantitative-finance practitioners
- +2 more
AI Tools by Chelis Lang
(1)
Chelis
Open Source Numerical Computing Language
Discussions
No discussions yet
Be the first to start a discussion about Chelis Lang
Latest News
Chelis repository continued active development; latest commit fixed owned Deep Expr teardown stack overflow.
Chelis 0.19.1 released with GitHub-based public toolchain/package installation, checker and Reef fixes, and newline-delimited JSON MCP support.
Chelis publicly launched as a tensor language agents can write and prove.
Products & Services
A statically typed functional tensor language and compiler for numerical programs written by AI agents. Surf (.ch) is the readable syntax and Deep (.dp) is the canonical representation; programs can be checked, evaluated, proved, and compiled to native executables.
A numerical-methods, statistics, and optimization shell covering special functions, probability distributions, linear algebra, statistics, root and differential-equation solvers, integration, and optimization.
A typed-dataframe shell with joins, group-bys, and rolling windows.
A quantitative-finance shell including option pricing, Greeks through automatic differentiation, volatility surfaces, yield curves, day counts, XVA, and currency-tagged money values.
Market Position
Chelis positions itself as a verification- and compiler-first alternative for numerical programs that agents write, contrasting its compile-time shape, precision, effect, ownership, and proof checks with Python/NumPy-style workflows where plausible numerical errors can survive tests. Its differentiators are tensor dimensions and precision in types, agent-oriented JSON diagnostics/MCP tooling, explicit reproducible randomness, and proof/property checking.
Leadership
Executive Team
Jeff Smith
Core team member
Named by the official launch post as a member of the Chelis core team; the same post says he spent years working on PyTorch.
Robert Ronan
Core team member
Named by the official launch post as a member of the Chelis core team; he is also the GitHub author shown on recent Chelis repository commits.
Founding Story
Chelis was started because AI agents increasingly write software that people review, but numerical mistakes can look plausible: a transposed matrix remains a matrix and an incorrect probability still prints as a number. The initial vision was a language designed around the agent loop of write, check, read diagnostics, and fix, with tensor shapes and element types in the type system and a prover that supplies evidence about stated properties.
Business Model
Revenue Model
Chelis is MIT-licensed open-source software. The official site says it is released by C Proof and identifies commercial support from C Proof; no paid Chelis pricing tiers were listed in the reviewed sources.
Target Markets
- AI coding-agent and developer-tooling teams
- Researchers and engineers doing numerical computing
- Scientific-computing users
- Quantitative-finance practitioners
- Data-analysis and typed-dataframe users
- Economics and dynamic-programming modelers
- Numerical code generated by AI coding agents and reviewed by people
- Scientific computing and statistics
- Quantitative finance and options/Greeks modeling
- Typed dataframe analysis
- Economic and dynamic-programming models
- Reproducible numerical programs where shape, precision, and correctness evidence matter