Skip to content
Try Hanzo
Language: English (United States)

Describe the change. Dev ships it.

Hanzo Dev is a coding agent. It reads your repository, makes the change, runs the tests and shows you the diff — in your terminal, on the web, in the cloud and in Slack.

Hanzo Dev · Coding agent
$ curl -fsSL https://hanzo.sh | sh
$ dev
v0.7.7 · Run /help for commands
┃ Describe the change.
✓ Reads the repository, makes the change and runs the tests
✓ Shows you the diff and waits for your review
src/retry.ts+1 −1 M
  • @@ src/retry.ts
  • export async function retry(fn, tries = 3) {
  • - for (let i = 1; i < tries; i++) {
  • + for (let i = 0; i < tries; i++) {
  • try { return await fn(); } catch (e) { last = e; }
  • }
auto·enso-autoEnso by default, every Hanzo model by name@acme/api (main)

3.02× faster on the same coding tasks.

Hanzo Dev and Claude Code were handed the same 44 coding tasks (single-fault repairs), one try each. Each clock replays that side's median time a task, at the speed it ran.

  • Claude Code: 9.52 s median time a task, 3.8 model calls a task
  • Hanzo Dev: 2.29 s median time a task, 1.0 model calls a task

3.02× faster over the 42 tasks both solved, 95% interval 2.06–4.36×.

Claude Code

done

9.52 s

median time a task

model calls a task3.8

replay of run coding.repair-held-devfix-claude-code

Hanzo Dev

done

2.29 s

median time a task

model calls a task1.0

replay of run coding.repair-held-devfix-hanzo-dev-dc

By how Hanzo Dev settled a task:

  • Settled by search, no model call, 24 tasks8.22×· 95% 6.86–9.82×
  • The model’s first edit, 4 tasks1.31×· 95% 1.14–1.48×
Every coding task and both runs →

One agent

Wherever the work is.

The same agent, the same models and the same account in every place it runs.

In your terminal, on the web, in the cloud and in Slack

  • Terminal

    dev

    In any repository. It reads the code, makes the change and runs commands in a sandbox. Bare hanzo opens the same session.

  • Web

    hanzo.ai/dev

    Describe an app. Dev writes it, runs it and deploys it, with the database, the sign-in and the storage already there.

  • Cloud

    dev cloud exec

    Hand a task to Hanzo’s cloud and close the laptop. dev cloud diff shows what it did; dev cloud apply puts it in your checkout.

  • Slack

    An org admin connects the workspace in Dev’s settings, and the team talks to it from their channels.

Install

One line, three ways.

Linux, macOS and Windows. Dev updates itself the way it was installed.

hanzo.sh
curl -fsSL https://hanzo.sh | sh

Linux and macOS. One native binary, its checksum verified.

Then sign in. Hanzo IAM links the machine to your account.

bash
dev login                 # pick an account: Hanzo is the default
dev login --hanzo         # straight to your Hanzo account
dev login --with-api-key  # or paste a Hanzo API key, hidden

Enso by default. Any model by name.

Dev starts on enso-auto through api.hanzo.ai, so every call is metered on your Hanzo account. Every model Hanzo serves is a flag or a /model away.

01Models

bashenso-auto

Kai decides. Dev does.

Kai is Dev’s own controller, not an adviser. At each step it returns an operation Dev can run, or says plainly that it abstains. The cheapest step that holds is the one taken.

02Controller

One candidate, its preconditions hold: it runs, and no model is asked.

Built to be checked.

Nothing counts as done because a model said so.

03Proof

  • Verified, not predicted

    A step is settled by what the verifier records — the exit status, the tests, the diff — never by what anything expected.

  • Bound to the revision

    Preconditions are checked just before an action runs, and an edit voids every decision made against the files as they were.

  • A ledger you can replay

    Every decision, fallback, run and verification is a line with a stable id and its reason, credentials redacted, so its time and cost can be read back.

Inside the fence you set.

Commands run in a sandbox and are approved inside it. Ask before every action, or lift the fence for one run. A repository’s own tool servers stay off until you say you trust it.

04Sandbox

Dev is our agent. Gen runs any agent.

Hanzo Gen is the harness: Claude Code, Codex, Cursor and the rest, on Hanzo’s tools and models and linked to your account. Dev is the agent it opens when none is named.

05Gen

Open source

Read every line.

Dev is Apache-2.0 on GitHub. Build it, change it, send a pull request — hanzoai/dev.

Questions

Questions & Answers

  • Dev is free and open source. Its model calls go through api.hanzo.ai and are metered on your Hanzo account at the price list’s rates, so one account and one bill cover every model.
  • Linux, macOS and Windows. hanzo.sh installs it on Linux and macOS, npm everywhere, Homebrew on macOS and Linux.
  • The way it was installed: from its own release, from npm, or from Homebrew.
  • Yes. dev login --with-api-key takes a Hanzo API key, hidden as it is typed or read from piped stdin.
  • Dev is Hanzo’s coding agent. Gen is the harness that runs any coding agent, Dev included, on Hanzo’s tools and models.
The agent, its controller, its source.Where to start, what Dev thinks with, and the code.

Read

Paste and ship

Paste this into any agent. It reads the skill manifest and calls Dev from there.

Agent prompt
Read https://hanzo.ai/skill.md and use Hanzo Dev in my project. Start with: dev

Describe the change. Dev ships it.