agentlanguages.dev
§ as of 20 may 2026 · 21 languages tracked

Programming languages designed for AI agents to write.

A live catalogue of an emerging field. 21 projects, three camps, and a foundational argument about what the problem even is.

Originally catalogued in a blog post “Three camps alike in dignity” for Negroni Venture Studios.

§ 01 · the taxonomy

Three camps disagree about what the problem is.

The Syntactic camp says the problem is representational — strip ambiguity at the token level. The Verification camp says it’s semantic — make contracts mechanically checkable. The Orchestration camp says it isn’t a language problem at all — constrain how agents coordinate. The catalogue below treats them as evidence that this disagreement is real.

§ 02 · the representational camp

Syntactic.

If the problem is that models trip on syntax, the fix is to strip ambiguity from the syntax itself.

The syntactic camp treats the problem as one of representation. Models choke on tokens that mean different things in different positions, on operators that need disambiguation, on whitespace that might or might not be load-bearing. Their answer: build a syntax where every token has one job.

The strongest entries replace text with structure (X07’s JSON ASTs), eliminate operators in favour of keywords (NERD), or surface intermediate representations as the user-facing form (Magpie’s SSA). The weakest are exercises in extreme density — one-character opcodes — that read more as conceptual art than as production languages.

What unites them is a belief that the LLM’s job is easier if the surface is simpler. What divides them from the verification camp is the absence of any mechanism for the compiler to catch what the model gets wrong.

§ 03 · the semantic camp

Verification.

The model doesn’t need to be right. It needs to be checkable.

The verification camp accepts that LLMs will keep making semantic errors and asks a different question: can the compiler catch them? Their answer is mandatory contracts, refinement types, effect systems, and SMT-backed proofs — the machinery of formal methods, repurposed as a guardrail for generative code.

This is the camp with the most established theoretical foundation (the work goes back to ML, Coq, Dafny, and Lean) and also the most mature implementations. MoonBit ships a full toolchain with ICSE-published research; Vera ships measured benchmark wins against Python on zero training data; AILANG claims to be written autonomously by its own coordinator.

The provocative entry is Prove — same diagnosis, opposite conclusion. It uses verification to make the language resistant to AI generation. The licence explicitly prohibits training use. Same camp, opposite politics.

§ 04 · the coordination camp

Orchestration.

It isn’t a language problem. It’s an agent-coordination problem.

The orchestration camp re-frames the question. The trouble with LLM-authored code, they argue, isn’t any specific defect in the code — it’s that agents need to be sequenced, sandboxed, audited, and approved by humans at the right points. The language is just the substrate; the runtime is where the action is.

Some entries here are academic (Pel from CMU, Quasar from Penn); some are infrastructure (Marsha treats the LLM as the compiler itself); one is a serious engineering effort aimed squarely at regulated industries (Boruna’s hash-chained evidence bundles and deterministic replay).

This camp overlaps the most with the others — Boruna includes a type checker and capability system; Quasar adds conformal prediction. The line between “language for agents” and “framework for agents” is genuinely blurry here, which is itself informative.

§ 05 · the full catalogue

The full set.

Every language designed for LLMs or agents to author code, as of 20 may 2026.

Camp
Sort
Syntactic

B-IR

“Maybe the most LLM-optimized language had been inside us the whole time.” A blog-post exploration, not a project.

Jason Hall (Chainguard)

Python (bootstrap) · Arm64 assembly

★ — N/A mar 2026 thought experiment
Syntactic

Laze

Minimal indentation-based syntax, no punctuation. Compiles to C. Optimised for generation speed, not correctness.

kerv

Python · C (via gcc/clang)

★ 3 Unknown apr 2026 early
Syntactic

Magpie

SSA as the surface syntax. Every value %-prefixed and typed at definition; one canonical way to express each operation; compiles to native via LLVM.

Magpie Language Developers

Rust · LLVM IR / native, also WebAssembly

★ 53 MIT apr 2026 early
read analysis →
Syntactic

NERD

Replaces all operators with English keywords. Built-in MCP client operations. Claims 50–67% token savings vs JS/TS.

Guru Sattanathan

C · LLVM IR

★ 135 Unknown jan 2026 working compiler
Syntactic

Sever

Single-character opcodes for extreme density. Explicitly a thought experiment; everything Claude-generated.

Avital Tamir

Claude-generated · Zig-based native (claimed)

★ 68 Unknown feb 2026 thought experiment
Syntactic

X07

Eliminates text syntax entirely. Programs are JSON ASTs; edits are RFC 6902 JSON Patch operations.

Author unknown

Unknown · WebAssembly

★ 7 Unknown apr 2026 early
Verification

AILANG

Purely functional, effect-typed, deterministic. Hindley-Milner with row polymorphism. Capability-based effects. No loops.

sunholo-data (Mark Edmondson)

Go · Native executables

★ 31 Apache 2.0 jan 2026 working compiler
Verification

Aver

Every function carries intent, declared effects, and a colocated verify block. Pure verify blocks export to Lean 4 theorems and Dafny lemmas; effectful ones lift through the Oracle proof export.

jasisz

Rust · bytecode VM, Rust, WebAssembly GC (Lean 4 / Dafny via proof export)

★ 43 MIT mar 2026 working compiler
read analysis →
Verification

MoonBit

AI-friendly general-purpose language. ICSE 2024 paper on real-time semantics-aware token sampling. Three years of training data.

Hongbo Zhang / IDEA Shenzhen

OCaml · WASM GC, JavaScript, native (C codegen), LLVM

★ — Unknown jan 2023 working compiler
read analysis →
Verification

Pact

Intent blocks, pipeline syntax, MCP server with five tools, LSP server. Close to Aver in design DNA.

Viktor Kikot

Rust · Interpreted

★ 1 Unknown apr 2026 early
Verification

Prove

Intent-first language with verb-based IO, refinement types, and contracts. Source is covered by the Prove Source License v1.0, which prohibits use as AI training data.

Magnus Knutas

Python (bootstrap) · C (then native via gcc/clang)

★ — Prove Source License v1.0 (language & .prv source) / Apache-2.0 (tooling) feb 2026 working compiler
read analysis →
Verification

Raskell

Not a new language — fix Haskell's tooling. hx replaces cabal/stack/ghcup; BHC is a clean-slate Haskell 2026 compiler.

Raffael Schneider

Rust · LLVM (via BHC)

★ — Unknown apr 2026 early
Verification + Orchestration

Vera

Mandatory contracts on every function. Z3 SMT verification. Typed slot references replace variable names. LLM inference is a first-class typed effect.

Alasdair Allan

Python · WebAssembly

★ 352 MIT feb 2026 working compiler
read analysis →
Verification + Syntactic

Zero

Agent-first learnability. “One obvious way to express most things.” Structured JSON output from every CLI command.

Vercel Labs

C · Native executables (linux-musl-x64)

★ 3.8k+ Apache 2.0 apr 2026 working compiler
Orchestration + Verification

Boruna

Deterministic, capability-safe workflow execution. Every effect declared, policy-gated. Hash-chained tamper-evident evidence bundles.

escapeboy

Rust · Bytecode (custom VM)

★ 0 MIT apr 2026 working compiler
read analysis →
Orchestration

Marsha

LLM as compiler backend. English specs + type declarations + I/O examples produce tested Python. The model IS the compiler.

David Ellis (Alan Technologies)

TypeScript · Python (via LLM)

★ 469 Unknown jan 2023 early
Orchestration

Pel

Lisp-inspired language for orchestrating AI agents. Grammar-level capability control. Natural language conditions.

Behnam Mohammadi (CMU)

Unknown · Unknown

★ — N/A (academic paper) may 2025 research paper
Orchestration + Verification

Quasar

Transpiles a Python subset. Automated parallelization, conformal prediction, user-approval security. 42% time reduction on ViperGPT.

Stephen Mell et al. (Penn)

Unknown · Custom runtime (from Python subset)

★ — N/A (academic paper) jun 2025 research paper
§ 06 · adjacent & unclassified

Related work.

Infrastructure that operates around these languages, or candidates that haven’t shipped enough to classify.

Adjacent / infrastructure

Plumbing

Typed coordination language for AI agent systems, built on applied category theory. Defines the wiring orchestration languages run within.

William Waites (Edinburgh / Leith Document Company)

Unknown · Native (binary downloads for macOS, Linux)

★ — Unknown mar 2026 working compiler
Unclassified + Orchestration

Spec

Two-domain architecture separating semantic specification from language-specific implementation. v0.2 draft, mostly unimplemented.

M. Abdullah Onus

Unknown · Unknown

★ 7 Unknown apr 2026 thought experiment
Unclassified

Valea

“AI-native systems programming language.” Announced on Hacker News. Too little public information to classify.

Hans Voetsch (Google)

Unknown · Unknown

★ 3 Unknown mar 2026 early
§ 07 · about this catalogue

Methodology.

A community-edited catalogue of programming languages designed for AI agents to author code — not languages that use LLMs at runtime, but languages whose syntax, semantics, or runtime are deliberately shaped around an agent being the primary author.

What counts

The bar is intent. A language is in scope if its designers explicitly target LLMs or agents as authors — through token-friendly syntax, mechanically checkable contracts, agent-coordination primitives, or first-class effect declarations for model calls.

A tool that calls an LLM at runtime (chatbots, autocomplete plug-ins, IDE assistants) is not in scope. A language whose only innovation is “works with Copilot” is not in scope. The line is the design intent, not the tooling.

How to contribute

Open a pull request adding a Markdown file to src/content/languages/. The contribution guide asks for evidence the project meets the inclusion bar and a self-classified camp with justification.

The maintainer reviews each submission for fit, accuracy, and tone (the catalogue is descriptive, not promotional). Marginal cases get discussed in the PR thread. Edits to existing entries are welcome — especially corrections from the language’s authors.