A managed cloud your agents drive: workstations, worktrees, databases, and compute under one control plane. Claude Code, Codex, DOSSBOT, or your own harness all speak the same API.
Reference designs
Real workloads we've built for. Pick one to see the cost breakdown. Numbers come straight from the same pricing matrix your bill runs on, no marketing fudge.
One supervisor remoco runs your pipeline orchestration, retries failures, reports status.
Agent spins up on demand per PR, runs the diff through your review harness, posts comments.
One standard box, 40 clock-hours, Lite tier -- a comfortable start.
Every engineer gets a large remoco for the shared dev loop, N worktrees per person, shared cache, idle-suspend off-hours.
| Line item | Qty | Cost |
|---|---|---|
| Box base | — | — |
| Clock-hours | — | — |
| Model tokens | — | — |
| Network egress | — | — |
| GPU hours | — | — |
| $200 grant | — | — |
Product
Remoco runs your dev loop on a cloud workstation that suspends when idle and wakes in seconds. Your URLs are stable, named, and access-controlled. Your agents operate directly on the workstation. Your org admin sees the whole fleet in one place.
VM-backed, persistent home disk. Suspend at 6h idle, resume in ~10s. Pre-baked image has Claude Code, cloudflared, portless, doppler, bun, docker.
Read →Every worktree, service, and user gets a stable subdomain. main.web.wiley.remoco.dev. Same shape internal and external.
Toggle between public and private with one flag. Private URLs are gated by WorkOS SSO; your team signs in with Google or Okta.
Read →One command spins up a branch in its own worktree with its own docker stack. Three feature branches side-by-side, no port collisions, each with its own live URL.
Read →A pill on every dev page shows branch, service, memory, containers. Click to expand into the full workstation dashboard. Zero app-side integration.
Read →Works with Claude Code, Codex, DOSSBOT, and your own harness. Every remoco command is invokable from the agent directly; agents drive the stack.
Read →Opinionated base images for Node, Python, Go, Rust. Derivative images per org. One click to add your AI harness or team tooling to the baseline.
GCP today, AWS next, Azure after. Same control plane, different backends. Pick the cloud per workspace or per org.
Bring your own API keys or route through us. Token spend attributed per user, per repo, per agent run. Budget caps and alerts built in.
How it works
No custom DNS to manage. No ngrok tokens. No K8s to reason about. Your workstation is addressable, gated, observable, the second you sign up.
Every URL starts gated. Toggle public with one flag; revoke with another. Immutable deployment versions snapshot a workspace for production traffic, so public URLs never point at a mutable dev server.
The URL shape is <branch>.<service>.<user>.<org>.remoco.dev. Internal localhost matches the same shape. Collab friction: zero.
The CLI, debugger, injector, and overlay are MIT-licensed on GitHub. The managed control plane is what you pay for: AAA, billing, shared observability, pooled capacity.
Managed control plane
Every workstation, every URL, every agent run, every dollar. The managed plane is where admins live; engineers rarely see it, but they're glad it's there.
| URL | Access | Owner | Last hit |
|---|---|---|---|
| main.web.wiley.doss.remoco.dev | public | wiley | 2m ago |
| hero.tetris.wiley.doss.remoco.dev | private | wiley | 18m ago |
| main.api.jaylen.doss.remoco.dev | private | jaylen | 1h ago |
| qa.web.ronil.doss.remoco.dev | public | ronil | 3h ago |
| main.docuforge.daman.doss.remoco.dev | private | daman | idle |
Two ways in
Drive the whole setup from the agent you already use, or let DOSSBOT walk you through it.
Get started
No YAML. No Terraform. No DNS console. Sign in with GitHub, pick a machine, start coding.
GCP by default. Pick a machine class or let us auto-size based on your repo.
Pick one or more repos from your GitHub orgs. We clone + install on first boot.
One-click expose. Stable URL under your-name.remoco.dev, public or private.
Pricing
Start with $200 in credits. Pay for what runs: box base + clock-hours + model tokens. Bring your own API key and pay compute only.
+ $0.40 per clock-hour active
+ $0.75 per clock-hour active
+ $1.40 per clock-hour active
Choose Heavy, Balanced, or Lite per session. Bring your own API key (BYOK) and pay $0 for tokens -- just the box rate.
| Model tier | Use case | $/1M input | $/1M output |
|---|---|---|---|
| Heavy | Complex reasoning, long context | $6.50 | $32.50 |
| Balanced | General coding and analysis | $3.90 | $19.50 |
| Lite | Fast tasks, triage, summaries | $1.30 | $6.50 |
| BYOK | Your own API key | $0 | $0 |
Runtime classes
Your home disk persists across size changes. Suspend-resume takes ~10s. Pay only for hours your workstation is actively running, not the storage it leaves behind.
| Class | vCPU | RAM | Disk | GPU | $/hour active | Best for |
|---|---|---|---|---|---|---|
smallweb · small api |
4 | 16 GB | 100 GB | — | $0.18 | Web + small API |
standardfullstack dev |
8 | 32 GB | 100 GB | — | $0.34 | Fullstack dev |
largemonorepos · parallel worktrees |
16 | 64 GB | 200 GB | — | $0.62 | Monorepos, parallel worktrees |
highmemdocker · elasticsearch · clickhouse |
8 | 64 GB | 200 GB | — | $0.58 | Heavy docker / ES / clickhouse |
gpu-l4inference · model eval |
8 | 32 GB | 200 GB | L4 × 1 | $1.04 | Inference, model eval |
gpu-a100training · fine-tunes |
12 | 85 GB | 200 GB | A100 × 1 | $3.90 | Training + fine-tunes |
$200 in credits on signup. Pay for what runs: box base + clock-hours + model tokens.