Topics / #typesafe

Typesafe projects on GitHub

Type-safe utilities and frameworks across languages.

As of , the most-starred Typesafe projects on GitHub are create-t3-app (29.2K stars), lo (21.4K stars) and open-saas (16.1K stars). GitHub.Locker tracks 14 Typesafe repositories with 89.6K stars combined. The fastest-growing on October 16 was awesome-jev-live, with +12 stars in a day.

14Projects
89.6KStars in total
+60Stars in a day

Hot right now

Most stars gained in a day · Oct 16

The all-time top

14 projects · by stars

  1. #1 · t3-oss/create-t3-app

    create-t3-app

    create-t3-app is an interactive CLI that scaffolds a full-stack, typesafe Next.js app built on the T3 Stack — a web dev stack focused on simplicity, modularity, and full-stack typesafety.

    29.2K
  2. #2 · samber/lo

    lo

    samber/lo is a Lodash-style utility library for Go built on 1.18+ generics: dozens of helpers for filtering, mapping, grouping and reducing slices, maps, strings, channels and functions.

    21.4K
  3. #3 · wasp-lang/open-saas

    open-saas

    Open SaaS is a free, open-source SaaS boilerplate built on the Wasp full-stack framework (React, NodeJS, Prisma). One scaffold gives you auth, payments, email, background jobs, a landing page, file uploads, an admin dashboard and more out of the box.

    16.1K
  1. 4

    router

    15.2KTanStack/router · TypeScript

    TanStack Router is a modern router for the web built around end-to-end type safety — routes, params, loaders and search params are all checked. The repo also hosts TanStack Start, a full-stack framework built on top of the router.

  2. 5

    skills

    2.7Ktypesafe-ai/skills

    A set of agent skills for building with TypeSafe: they let your coding agent get typed decisions and probabilities from System One models.

  3. 6

    typesafe-computer-use

    1.2Kawlevin/typesafe-computer-use · Python

    A Python agent that drives your Mac toward a goal you type in plain English: it OCRs the screen, a cheap classifier picks each action, and it clicks — about $0.0002 per step.

  4. 7

    awesome-jev

    975heyjunpenn/awesome-jev · Astro

    A verified, community-maintained catalog of open-source projects built with Jev — TypeSafe AI's System One model for typed decisions inside software. The README snapshot covers 832 unique public GitHub repositories.

  5. 8

    von

    916wfzyx/von · Python

    Von is an open-source, non-autoregressive decision model (395M parameters, 1.5 GB) that makes calibrated discrete, probabilistic, and ordinal decisions in a single forward pass in under 25 ms — a local, free alternative to the proprietary TypeSafe Jev.

  6. 9

    foreman

    771thruwire/foreman · Python

    Foreman is a supervisor that watches a coding agent (like Codex or OpenCode) while it works on a ticket or bug report, using a fast decision model called Jev to independently judge whether the implementation is complete, needs verification, or needs a human.

  7. 10

    jev-experiments

    406dabit3/jev-experiments · TypeScript

    A monorepo of latency-focused demo apps built around TypeSafe / Jev, each of them written by the Devin AI coding agent.

  8. 11

    awesome-jev-live

    338wh000wh000/awesome-jev-live · Python

    Awesome Jev Live is an evidence-graded index of projects built around Jev, TypeSafe AI's System One model — 836 entries covering SDKs, MCP tools, agents, apps and open models, published in 20 languages.

  9. 12

    jevmem

    127Avinash-jetwani/jevmem · TypeScript

    jevmem saves the decisions, rules, and failed approaches from your Claude Code (and other coding agent) chats into a JEVMEM.md file in your repo, and brings the relevant ones back in later sessions so context survives a fresh chat.

  10. 13

    Bud-Decision-Studio

    110BudEcosystem/Bud-Decision-Studio · HTML

    Bud Decision Studio is a cross-platform desktop app and local serving system for 'decision models' (Jev-like 'System One' models): you give it a situation and typed questions, and it returns a calibrated probability for every allowed answer instead of generated text.

  11. 14

    jevals

    105openlayer-ai/jevals · Python

    jevals is a Python library for evaluating and guardrailing AI agent traces using 'Jev' decision models instead of a text-generating LLM judge. It bundles every check for a trace into one request that costs a fraction of a cent and returns in a few hundred milliseconds, so you can eval every trace and even run checks inside the agent loop.

Often tagged together