Prepare a repo
Clone a repository, inspect files, install dependencies, run tests, and return a concrete status without touching the user machine.
Cloud for Agents
A cloud built around agent work
Agents need more than a prompt and a model call. They need a place to run, a way to keep state, a boundary for approvals, and somewhere to publish what they produced.
Clanker Cloud sandboxes give agents a hosted workspace for commands, repo work, tool calls, workflow records, traces, handoff memory, and hosted outputs.
A useful agent run needs a runtime, state, audit trail, approval gates, and an output. Clanker Cloud makes those pieces part of the sandbox.
Agent work
What agents can do in Clanker Cloud
Run a workflow
Create a workflow record, append traces, store handoff memory, and pause before side effects.
Publish a result
Use a paid account to publish a static report, review page, preview app, or run status from the sandbox.
Hand off cleanly
Keep enough non-secret context in memory and traces for the next agent to continue without rereading the whole task.
Connect to DevOps
Use Clanker DevOps and MCP when the agent needs live infrastructure context or reviewed production action.
Automate repeat work
Turn stable agent runs into recurring workflows hosted on Clanker Cloud or in a private enterprise environment.
Sandbox runtime
Create hosted sandboxes for agents that need a shell, workspace, and clean execution boundary.
Workflow state
Record the job, steps, traces, approvals, memory, and artifacts with the sandbox instead of scattering state across prompts.
Human gates
Ask for approval before publishing, deleting, sending, installing, changing accounts, or touching production systems.
Hosted outputs
Paid accounts can publish static reports, previews, and status pages from the sandbox.
Supported providers
Works across the environments teams already run
The current product positioning covers cloud providers, Kubernetes, GitHub, and bring-your-own AI keys from one local operating surface.
Supports ->GitHubKubernetesBYOK
Sandbox tiers
Where each sandbox type fits
| Caller | What it gets | Best for |
|---|---|---|
| Anonymous | One-hour sandbox with a sandbox token returned at create time. | Trying the API, short commands, demos, and quick agent tests. |
| Free account | Account-owned sandbox that lasts about eight hours. | Normal work sessions, repo checks, and agent experiments. |
| Paid account | Larger persistent sandbox with hosted sites and /explain. | Long-running workflows, team runs, reports, previews, and heavier agent environments. |
Build with it
Start from the API or the product surface
FAQ
Common questions
Is this only for coding agents?
No. Coding agents are a common starting point, but the same sandbox model works for research workers, office automation workers, DevOps helpers, and custom internal agents.
Where should secrets live?
Do not put secrets in sandbox memory or trace records. Use the user account, Clanker Cloud settings, the desktop boundary, or the customer-controlled secret path for private deployments.
Next step
Create the first sandbox
Use the sandbox API for a hosted runtime, then add workflow records and approvals when the run becomes more than a one-off command.