Skip to main content

Clients & tools

Zeq is a language, and these are the ways to speak it. Every client — a one-line CLI, a typed SDK, a browser extension, an AI tool surface — wraps the same wire calls, shares the same constants, and gets back the same Ed25519-signed, independently-verifiable result. Same equations, same bits, from any of them. Pick the surface that fits where you work; the physics underneath is identical.

The ways in​

Code

From your terminal, your app, or any language.

  • Terminal CLI — one line to install, zeq tutorial to learn
  • Language SDKs — TypeScript/JS, Python, and more; offline-verifiable

Connect an AI

One optional avenue — give a model the rules and the instruments.

  • MCP server — the Zeq engine as callable tools for any agent
  • Kernel skill — the kernel as a SKILL.md an agent reads
  • The AI harness — the whole surface an AI consumer works through

In your tools

The VM inside the tools you already use.

  • VS Code & Chrome — the ZeqVM in your editor and browser
  • ZeqGit — the framework's own git + PDF repository

One identity, everywhere — and one machine, in one place​

Your ZeqID is the same on every surface and on every node. It is derived from your account equation by computation, not handed out by a server, so all eighteen nodes compute the same ZEQ7… from the same input. Sign in from the terminal, the editor, the browser or the web and you are the same identity.

Your credits follow you too. The compute-credit balance is not per-node: hulyas.org is the credit authority, and every other node is a thin reader that forwards your balance and conversion requests to it over the signed Ed25519 peer protocol. There is exactly one credit balance for a ZID across the whole network — see the economy at a glance. (That routing is configured per node with a vault secret; we verified the mechanism in the source, not the setting on each of the seventeen forks.)

Your machine does not follow you. Machines, contracts, envelopes, coins, pages and the observer feed live on the node that created them and are not replicated. A zsm_… key is scoped to one machine on one node, and each node issues its own session. So "the same machine in every client" holds only while every client points at the same node — which is the normal case, and is worth being deliberate about rather than assuming. If a client shows you an empty account, you are almost certainly pointed at a node that does not home you: ask GET /api/identity/home/:zid before you create anything new.

Every surface offers Create Account on both tiers, not just sign-in, and creating an account never silently logs you in: first you get a screen to save your generated equation and download your .ZEQ recovery file — there's no password reset, so that file is your only way back in. Three zero-friction methods:

  • Simple — email + four words; the client derives one strong equation.
  • Expert — your ZeqID + the equation you saved.
  • .ZEQ — upload your encrypted recovery file + PIN; no password typed.

Start with a hosted machine​

You don't install the framework — you spin up a machine (free) and it runs on one node of the eighteen: the node you spun it up on, which becomes its home. Each node is a whole machine in its own right — its own Postgres, its own Redis, its own clock — not a slice of a shared pool.

Once you have a key, use whichever client fits where you work. They all speak the same wire calls against the same constants, and a result signed on one node re-verifies on any other, because verification travels even though machines don't. Point your client's base URL at your home node and every surface sees the same machine. Reach for a client only if you want one — the machine runs with or without any of them installed.