CodeHerderSearch⌘KRequest access →

Get access, then ship your first task

Ask for access first. A person replies and sets up your workspace. Then pick a path. Use the web app with no code, or use the ch CLI from your terminal.

A developer? Try CodeHerder on your own machine first. You need no account. See the local trial →

Step 0 · Get access

First, get your workspace

Sign-up isn't open to everyone yet, so both paths below start the same way: tell us what you want to build, plus a bit about your team if you'd like to share it, so we can set your workspace up correctly from the start. A real person reads every request and replies, usually within a business day.

A request is easiest to act on when it says what you want shipped first, where the code lives, and whether anyone on your team already writes code. Not sure how to phrase it? Browse starter briefs for real, ready-to-copy examples, or run your own draft through the brief checker.

What to have ready on day 1

  • One repo on GitHub or GitLab.
  • One macOS or Linux machine that stays on while agents work.
  • One supported coding agent signed in on that machine. Use your own provider key or plan.
  • gh or glab signed in on that machine. The merge stage uses it.
  • No machine to spare? An owner or admin can launch a device in your own AWS account from the Devices page. See where agents run.

Not a developer? Send this list to the person who helps you on day 1.

Not sure your repo fits? Check will it work first.

Path A · Build without writing code

File the task. Watch the herd build it.

Everything after the one-time device setup happens in the browser: no CLI, no git, no code.

  1. 1. Sign in

    Once your workspace is set up (Step 0, above), sign in at app.codeherder.com. A fresh workspace shows a Start here checklist: connect a device, add a repo, create a task, run it.

  2. 2. Connect a device

    Connecting the first machine is a one-time technical setup. Pair with a technical teammate or your IT once, or see where agents run for a one-click option that needs no CLI. Day to day, you work from the web app. Once it's running, the Devices page flips to "connected" automatically. No refresh needed.

  3. 3. Add your repo

    Open Repos → New repo and give it a name and a git clone URL. CodeHerder is not a git host. Your agents work directly in your existing repository. Starting from a prototype instead of an empty repo? See /prototype-to-production for what to check first.

  4. 4. File a task in plain English

    Open Tasks → New task. Give it a title and describe what "done" looks like, in plain language. Run it past the brief checker first, or copy one of the starter briefs.

  5. 5. Watch it ship

    The task moves through plan → code → review → merge → verify → done on its own. Want to sign off first? Add an approval gate to a stage.

Once your first task ships, see the two-week rollout plan for what to hand over next.

Path B · Developer CLI quickstart

Same pipeline, driven from your terminal

Copy-paste steps straight from the docs.

  1. 1. Install ch

    Sign in at app.codeherder.com and open Install CLI (under "Help"), which auto-detects your platform and links the prebuilt binary.

    ch · install
    ❯ curl -fsSL https://app.codeherder.com/cli/install.sh | sh
    ❯ ch version
  2. 2. Bootstrap your account

    The first request to a fresh codeherder is unauthenticated: bootstrap solves a local proof-of-work challenge and emails a one-time code, no separate signup form needed. On the hosted service this step is invite-only right now too (Step 0 gets you access first). Running your own codeherder instance instead? Bootstrap works immediately, with no waiting. With no email configured, the server just writes the code to its log.

    ch · bootstrap
    ❯ ch human bootstrap --email you@example.com --display-name "You"
    # solves a local proof-of-work (~1s), emails a 6-digit code, prompts for it
  3. 3. Start a device

    ch device-server self-registers on first run: no separate register step, and no agent to create, assign, or approve.

    ch · device-server
    ❯ ch device-server
  4. 4. Add a repo

    The repo's own git-host CLI (gh for github.com, glab for gitlab.com) must already be authenticated on the device. The merge stage uses it to open the pull/merge request.

    ch · repo create
    ❯ ch repo create --name my-app --url git@github.com:you/my-app.git --branch main
  5. 5. Create your first task

    Creating the task starts an agent. You do not move the task yourself.

    ch · task create
    ❯ ch task create --title "add a health check endpoint" --type story
  6. 6. Watch it ship

    Follow the task from the terminal, or open it in the app and watch the same stages (plan, code, review, merge, verify, done) update live.

    ch · watch
    ❯ ch task show <taskId>
    ❯ ch activity --wait 30

Once your first task ships, see the two-week rollout plan for what to hand over next.

Bring your own provider key. Run it on your infra.

Point your devices at your own server instead of the hosted deploy. Every model call and every dollar of spend then stays on your hardware, under your own security policies.

See the self-hosting guide →

Want the bigger picture?

How it works walks through the whole platform: workspace setup, capability routing, the task lifecycle, live sessions. The FAQ answers common questions for both developers and non-technical builders. Every ch verb and flag used above is documented in the CLI reference, part of the full docs. Still sizing it up? Will it work for your stack answers that directly.

Try it on one repository for two weeks.

File a task. An agent builds it. A person reads every request. You see the cost of each task.

Try "pricing", "connect a device", or "who reviews the code"

↑↓ move · ↵ open · esc close