Your AI coding agents,
in the cloud, in minutes.
Connect your repo, bring your own API key, and put real coding agents to work on isolated cloud infrastructure. No local installs, no VMs, no DevOps.
Free to sign up · Bring your own keys — any provider, your account, your spend
- Claude Code
- OpenAI Codex
- z.ai
- MiniMax
- Ollama
-
Your key, your provider bill
Tokens bill to the provider account you already hold. No markup, no pooling, no resale.
-
Isolated per organisation
Each org's code runs in its own resource-capped container. One org never sees another's workspace.
-
Nothing runs in your browser
Agents execute on our infrastructure, streaming every tool call, edit and approval back to you.
What works today
Everything here is shipped,
not on a roadmap.
No waitlists, no preview tiers, no vapor. This is what you get when you sign in.
Agents that actually run
- Cloud agent chat
- A browser chat backed by a workspace container running real agent CLIs — tool activity and token usage stream back live.
- Your provider, not ours
- Claude Code, OpenAI Codex, z.ai, MiniMax, Ollama. Paste the key for the gateway you want.
- Repo in, cloud out
- Push a repository straight into a workspace. Edits, git state and uncommitted work survive container rebuilds.
You stay in control
- Approvals at the gate
- A mutating tool call does not run until a human answers. Git operations and destructive actions wait for you.
- Every turn on the record
- Tool calls, approvals, git operations and chat turns land in an audit trail with token and cost metering.
- Hardened workspaces
- Per-org containers with resource caps, isolated filesystems and no docker socket.
Built for real work
- Organisations and roles
- Owner, admin, member and viewer — onboard a team in minutes, not meetings.
- Real accounts, real MFA
- Google or email and password, with email MFA and trusted devices. Your session stays yours.
- A macOS app too
- A native Git workflow with signed approvals, alongside the browser platform.
How it works
From sign-in to shipped commit in three steps
-
01
Sign in
Google sign-in or email and password — with email MFA and trusted devices once you add them.
-
02
Start your cloud workspace
Push your repo into a hardened, resource-capped container. Paste an API key — any provider — and you're provisioned.
-
03
Give an agent a task
Watch it stream tool-for-tool, approve the risky bits, commit the result. Approve a push right from the chat.
Your first ten minutes
A setup wizard that does the setup, not just describes it
Sign up and you land on a guided checklist of the six things an organization needs before it can run an agent. The first three happen inside the wizard — it starts your workspace, connects GitHub, and takes your provider key. The other three are one click each, in the places where the machinery already lives.
Bring two things. A repository on GitHub, and an API key from any supported provider. No card, no seat, no minimum spend — you are billed for the container that runs a turn, at cost.
-
01 Start your workspace
A real Linux container on our hardware, booted for you. It suspends when idle and keeps every file.
-
02 Connect GitHub
One approval for the whole organization. Then pick a repository and watch it clone.
-
03 Bring your API key
Claude, OpenAI, z.ai, MiniMax or a local Ollama. Stored encrypted, billed to you at your provider's own prices.
-
04 Run your first turn
Describe the change in plain language. Every tool call streams back; anything risky stops and asks you first.
-
05 Publish an app
A hosted site on its own address, every deploy a version you can roll back. Or install a base and start from a whole stack.
-
06 Add a database
Postgres, MySQL or Redis beside your apps. Credentials are generated and revealed exactly once.
Watch, don't guess
Every agent turn,
on the record.
Agents don't run in your browser and they don't run in the dark. Each turn streams from our infrastructure — tool calls as they happen, file edits with diffs, approvals held at the gate, and token and cost metering on every exchange. The whole thing lands in your audit trail.
Security & isolation
Agents run in their own box. Not yours.
Every organization gets its own workspace containers on our infrastructure — capped, isolated and hardened by default.
-
Isolated containers
Resource-capped, isolated filesystems, no docker socket. One org never sees another's workspace.
-
Secrets stay secret
Your API key and tokens live in OS-level safe storage and are never logged, exported or resold.
-
Approvals at the gate
Destructive actions and git operations wait for a human. The desktop app adds native signed approvals.
-
BYO key economics
Your provider bill reflects your usage — whichever provider you pick — with no markup, no token pooling, no surprise spend.
Pricing
Free to sign up. You bring the AI.
Agent tokens bill to the provider account you already hold. EasyCoder charges for the container that runs your code, by the second — and idle containers suspend.
You pay your AI provider
Anthropic, OpenAI, z.ai, MiniMax and Ollama — paste a key for the gateway you want. Tokens bill straight to your provider account at their prices. We never see that spend and we don't mark it up.
$0 markup · $0 pooling · $0 resale
We charge for the container
- Workspace
- $0.020 / container-hour
- App
- $0.005 / container-hour
- Database
- $0.012 / container-hour
- Storage
- $0.025 / GB-month
Priced by the second, on the second. Idle workspaces suspend after a short timeout — idle time is free. Databases can opt into scale-to-zero. Full breakdown in your billing tab once you sign in.
Pro plan
optional$20 / member / month
- ·50 app-hours included per member
- ·50 GB-month storage included per member
- ·Overage billed at the rates above
- ·Viewers don't count as seats
Free already covers 720 app-hours and 2 GB-month storage — one app, all month, no card needed. Switch to Pro from your billing tab once your team is past the included hours.
Who it's for
Built with its own agents.
EasyCoder is in production at easycoder.dev, and the agents you read about on this page are the same ones we build it with — real turns, real containers, real audit trail. When teams start putting their names here, they will.
Questions, briefly answered
- Do I need my own API key?
- Yes — one for whichever provider your agents should run on: Anthropic, OpenAI, z.ai, MiniMax and more. Keys ride each turn over TLS and are never stored server-side, so your spend stays visible on your own provider bill and is never marked up. Ollama inside your workspace even runs keyless.
- Which agents can I run?
- Claude Code, OpenAI Codex, z.ai, MiniMax and Ollama, each running as the real CLI inside your workspace — not a reimplementation of it. You pick the provider and paste the key; tokens bill straight to your account.
- Where do the agents actually run?
- On our infrastructure — hardened, per-organization workspace containers with resource caps and isolated filesystems. Nothing installs on your machine.
- What happens to an idle workspace?
- It suspends after a short timeout and stops billing — idle time is free. Published apps and databases run continuously by default; databases can opt into scale-to-zero from their settings. Everything that suspends wakes back up the moment a request lands.
- What does EasyCoder itself charge?
- Two parts. First, the per-container usage at the card prices — workspaces $0.020/hour, apps $0.005/hour, databases $0.012/hour, persistent storage $0.025/GB-month. By the second, with no minimums. Second, an optional Pro plan at $20 per member per month with 50 app-hours and 50 GB-month storage included; overage still bills at the card prices. The Free tier already covers 720 app-hours and 2 GB-month storage, so one app can run all month without a card on file. Full rate card and your live usage are visible in the Billing tab once you sign in.
- What happens when I hit my included cap?
- Your running app workspaces pause, the Billing tab surfaces an "Included quota reached" banner, and an admin clicks Continue on overage to opt in and restart the paused apps. From that point on the org keeps running past the cap and bills at the card rates; the choice is sticky for the period. We do not silently bill overage — you have to say yes.
- Can an agent overwrite my repo without asking?
- No. Git operations and destructive tool calls sit behind approvals — you decide what makes it out of the workspace.
- Can I bring my whole team?
- Yes. Organizations carry owner, admin, member and viewer roles, with a shared audit trail and shared usage metering. Agents can also be assigned per-agent and addressed by name in a turn.
- Is there a desktop app?
- Yes — an EasyCoder macOS app with a native Git workflow and approvals, alongside the browser platform.