A community-edited catalogue of programming languages designed for AI agents to author code, the site is live at agentlanguages.dev.

The site catalogues an emerging field of language design that treats LLMs and autonomous agents — not humans — as the primary authors of code. As of May 2026, the catalogue tracks projects across three philosophical camps that disagree on how to frame the underlying problem:

• Syntactic — Strip ambiguity at the token level. Make syntax easier for LLMs to parse and generate.
• Verification — Make contracts mechanically checkable. The model doesn't need to be right; it needs to be checkable.
• Orchestration — Reframe the problem as agent coordination, not language design. Constrain how agents are composed rather than the code they produce.

Plus an Adjacent category for infrastructure that supports the field but isn't itself a language agents write (currently: Plumbing), and Unclassified for projects with insufficient public information to place.

The taxonomy originates in "Three camps alike in dignity" (Negroni Venture Studios, May 2026). This site is a community resource for tracking the field as it evolves — not a Negroni product.

The catalogue accepts pull requests for new languages, corrections, and substantive analysis. See CONTRIBUTING.md for the submission criteria, frontmatter schema, and review process.

The bar for inclusion: the project must be designed for LLMs/agents to author code — through token-friendly syntax, mechanically checkable contracts, agent-coordination primitives, first-class effect declarations for model calls, or similar. Tools that use LLMs at runtime (chatbots, autocomplete, code-completion plugins) are out of scope.

PRs are reviewed editorially by the maintainer. Not every submission is merged, but every submission gets a response.

Requirements: Node ≥ 22.12 and npm.

npm install
npm run dev      # local dev server at http://localhost:4321
npm run build    # production build to ./dist/
npm run preview  # preview the production build locally

A GitHub Action runs weekly (Monday 06:00 UTC) and refreshes src/data/stars.json from the GitHub API. The site does not make live API calls — stars are baked in at build time, and the footer surfaces the most recent refresh date. If the refresh fails for a given repo, the previous value is kept.

To refresh manually (e.g. after adding a new language entry and not wanting to wait until Monday):

./scripts/setup.sh                                                       # creates .venv/, installs scripts/requirements.txt
GITHUB_TOKEN=$(gh auth token) .venv/bin/python scripts/refresh_stars.py  # writes src/data/stars.json

Then commit src/data/stars.json. Homebrew Python on macOS refuses system-level pip installs (PEP 668), so the script's pyyaml dependency lives in a project-local .venv/ that scripts/setup.sh creates (and .gitignore ignores). The setup script is idempotent — safe to re-run.

GITHUB_TOKEN is optional; without it the script falls back to unauthenticated requests with a 60/hour rate limit, fine for the catalogue's size but easy to exhaust on repeated runs.

Maintained by Alasdair Allan. The catalogue grew out of research first published in "Three camps alike in dignity" for Negroni Venture Studios; this site is community-maintained, not a Negroni product.

• Code (Astro components, scripts, configuration): MIT
• Content (language entries, prose, analysis): CC BY 4.0

See LICENSE for full text of both.