Skip to content

Agents do the work.
You set the course.

Leidang runs your coding agents in the background, each on its own machine, with your own subscription. Claude Code and Codex today, OpenCode and Pi next. Start from your laptop, your phone, or Slack. Come back to finished work.

Session · running claude

Move the billing tests to Vitest and open a PR

You · keep the coverage threshold, open a draft when green

Claude Code · plan ready, three steps, waiting for you

Changes

  • vitest.config.tsadded
  • billing.test.tsmodified
  • jest.config.jsdeleted

Every task gets its own ship.

Describe the work. Leidang boots an isolated machine with the repository, the tools, and the credentials already aboard, and the agent starts. The transcript, plan, files, terminal, and preview stay with the work, not with your laptop. Close the lid; the ship keeps sailing.

The agent has a question

Two payment providers are configured. Which one should the new tests target?

Stripe onlyBothLet me explain

From your phone

Stripe only. Adyen is being removed next sprint.

Steer from wherever you are.

Approve a plan from your phone. Answer the agent's question from Slack. Start a task from the CLI and check on it from the web. One session, every door.

Web · CLI · Mobile · Slack · Schedules · Webhooks · API

Software that looks after itself.

Routines wake an agent on a schedule or an event: a failing check, a new Sentry issue, a review request on GitHub, a message in Slack. It reproduces the problem, fixes it, opens a draft pull request, and asks a human only when it has to.

The fleet goes out at night. You review in the morning.

  1. 02:14

    An alert arrives. A routine wakes an agent.

  2. 02:15

    A ship leaves the harbor with the repo aboard.

  3. 02:41

    Reproduced, fixed, draft pull request opened.

  4. 07:30

    You read it with your coffee and merge.

Bring your own harness.

Leidang is the runtime underneath, in the cloud or on your own infrastructure. It is not a model vendor's product and does not pretend to be. Bring the harness you already pay for, with the subscription you already have. No markup on tokens, no lock in. The whole platform is open source; run it yourself or use the hosted beta.

Agents
  • A wooden robot sailor standing in for Claude CodeClaude Codetoday
  • A wooden robot sailor standing in for CodexCodextoday
  • A wooden robot sailor standing in for OpenCodeOpenCodenext
  • A wooden robot sailor standing in for PiPinext

Adding one is a runner and a registration, not a rewrite.

Subscription

Yours. Connected once, used by every ship.

No markup on tokens. Leidang charges for the harbor.

Source

Open source under Apache 2.0.

Self host with Docker or the AWS stacks, or use app.leidang.dev.

Identity

Your SSO, your audit log.

Credentials encrypted and scoped to each workspace.

Your laptop was never where the work should live. Give every task a ship. Give your team one harbor.

Where this goes.

One runtime under every harness, every agent, every surface. The foundations ship today. The rest is stated as a direction, not a feature.

  1. A wooden robot sailor unrolling a harbor chart on a dock table

    Today

    Shared environments, skills, connectors, and sessions. Every ship leaves the harbor with the same charts.

  2. A wooden robot sailor writing in a ship's logbook by lantern light

    Next

    Agents that learn from outcomes: what got approved, what got reverted, what the reviewer changed by hand.

  3. A wooden robot lighthouse keeper polishing the lens in a lamp room lined with logbooks

    Eventually

    The runtime becomes the company's working memory. How you build, written down by doing, readable by people and agents alike.

A wooden robot sailor on a headland blowing a horn as small boats leave their coves and gather into one fleet

In the Norse leidang, every district kept a ship ready. When the call came, the ships left their harbors and joined one fleet under one command.

Three sails, one fleet. That is the whole product.

The practical answers.

Anything missing? Request beta access and the team will get you a real answer.

01What is Leidang?

The open source cloud runtime for coding agent harnesses. Claude Code and Codex run on it today, OpenCode and Pi next, each task on its own isolated machine, steered by your team from the web, your phone, Slack, the CLI, schedules, webhooks, or the API.

02Which agents can I bring?

Claude Code and Codex today, OpenCode and Pi next. Adding an agent is a runner and a registration, not a rewrite.

03Do I need a new subscription?

No. Connect the Claude or Codex plan you already pay for. Leidang charges for the harbor, not the tokens.

04Does the work keep running when I close my laptop?

Yes. The agent runs on its own machine, not in your terminal. Idle sessions sleep with their filesystem intact and wake on demand.

05What does software that looks after itself mean in practice?

Routines wake an agent on a schedule or an event. It fixes what it can, opens a draft pull request, and asks a human when it has to.

06Can I run it myself?

Yes. Leidang is open source under Apache 2.0 with a local Docker path and AWS stacks for production.

07How do private repositories work?

Through a per organization GitHub App with scoped installation tokens. It clones private repos, opens pull requests, and can start a session when you are requested as a reviewer.

08Is the hosted version open today?

It is an invite only beta. Request access and the team will follow up.

The four wooden robot sailors standing in a row on a sunlit dock with sailboats moored behind them

The fleet is ready.

Request beta access

For engineering teams that run coding agents every day.

By requesting access, you agree that Leidang may contact you about the beta.