A workstation for your army of coding agents.

Run swarms of agents for days at a time, together with your team, from software you'd happily sit in front of for twelve hours.

From repository to pull request in four steps.

Connect once, and every agent starts with your code, setup and secrets.

01

Connect a repository

Install the GitHub app and pick repositories. Each agent works on its own branch.

Repositories →
02

Import your setup

Bring your skills and MCP servers over from Claude Code, Codex or Cursor.

Importing →
03

Start an agent in a sandbox

Each cloud thread gets its own sandbox with your repo, secrets and setup script.

Sandboxes →
04

Verify and merge

It runs your tests and checks the change in a browser. You review and merge.

Verify and merge →

Everything a swarm of agents needs.

Agents that run for days

Hand off the long jobs. An agent keeps working through the night, fixes its own CI failures and answers review comments until the pull request is green.

How PR Auto-fix works →

Dozens of agents, one orchestrator

Give an orchestrator the whole job. It plans the work, sends out ranks of subagents in parallel, checks every one, and reports back.

Running threads →

Built for twelve-hour days

Hoplite is somewhere you work all day. It opens from your saved session, threads load as you reach for them, and your messages land the moment you send them.

Download for macOS →

Your whole team, in the same threads

Watch an agent work with your teammates, steer it together, and hand it off when you log off. Everyone sees the same thread, live.

Multiplayer agents →

Wired into your stack

Work starts where your team already talks. Mention Hoplite in Slack, assign it a Linear issue, or let a Sentry alert or a schedule start a fix while nobody's looking.

All integrations →

Close the lid.
Your agents keep going.

They run in the cloud for as long as the job takes, and message you when they need you.

6 agents running · 9h 41m

Secure by default.

Every thread runs walled off, your secrets stay encrypted, and Hoplite never trains on your code.

Isolated by default

Every cloud thread runs in its own sandbox, walled off from the rest.

Secrets stay sealed

Secrets are encrypted at rest and never shown again once saved.

Short-lived GitHub access

Agents get short-lived tokens, scoped to your project's repositories.

Never trained on your code

Hoplite doesn't train or fine-tune models on your code or content.

Questions

How is it different from Claude Code or Codex?
Hoplite runs them for you: in the cloud, for days, dozens at once.
Can I use my ChatGPT or Claude subscription?
Bring your Claude or ChatGPT subscription to run Claude Code or Codex in Hoplite.
Do I need to keep my laptop open?
No. Agents run in the cloud and message you when they need you.
Is my code safe?
Every thread runs in its own isolated sandbox, with short-lived GitHub tokens and encrypted secrets.
Does it work with Linear, Slack and Sentry?
Yes. Start work from any of them, and follow it in Slack and Linear as it lands.
What does it cost?
Free to start. Paid plans add more included usage per seat.

Deploy an army.
Free to start.

Bring your own Claude or ChatGPT subscription, or use ours.