Coding agents
that keep working.
While you’re away.
Your native coding harnesses, running in the cloud.
One workspace from the first prompt to the final diff.
The agent can code.
The workspace still needs you.
Delegating a task should not mean babysitting a laptop, rebuilding an environment, or piecing the result back together.
Work is tied to your machine.
A local terminal is a fragile home for a long-running task.
The work is scattered.
Conversation, files, running apps and review live in separate places.
Switching agents means more glue.
Different harnesses need explicit briefs, shared files and clear boundaries.
Statefold makes the workspace the durable part—not your terminal session.
A real workspace.
Not just a chat window.
Start in the browser or offload a local checkout. Come back to the conversation, files, terminal, preview and diff.
Different agents.
One body of work.
Claude Code
Explore the repository. Write the brief.
plan.mdCodex
Read the plan. Implement and test.
code + test resultsOpenCode
Inspect the diff. Leave actionable feedback.
review.mdThe harness owns the conversation.
Statefold owns the workspace.
Keep native tools, questions and permission requests. Connect your own providers instead of adopting a replacement agent.
Agents can delegate tasks and exchange briefs. Saved workflows order stages and use files for handoffs—not transferred conversation state.
Illustrative workflow. Harness support and authentication vary by provider. Codex and OpenCode lead the private-alpha rollout.
Delegate the task.
Keep the judgment.
Our initial focus
Independent developers and small engineering teams already using coding agents.
Bug fixes, full-stack features, browser checks and review work that can run without holding a local machine open.
What makes Statefold different
Bring the harness you trust
Native agents and user-owned provider connections.
Keep the whole working environment
Conversation, filesystem, terminal, preview and changes together.
Let agents work across boundaries
Explicit delegation, file-based handoffs and human review.
The unit of value is a task you can review and ship, not another answer in a chat.
Charge for the workspace.
Let builders choose the model.
Workspace subscription
Managed tasks, project context, collaboration and review.
Metered cloud compute
Execution resources for active tasks, with clear usage boundaries.
Users bring their own model-provider connections. We focus on the environment and the workflow around the harness.
Validate before scaling. Start with individual builders, learn from real task workloads, then expand into repeatable team workflows.
Proposed model, not current pricing. Commercial packaging and unit economics are still being validated.
A stronger foundation
for autonomous work.
We’re evaluating Daytona as an execution backend for Statefold’s cloud workspaces.
Startup credits would fund integration testing and early reliability validation—not GPU training.
- Initial test profile
- 2 vCPU / 8 GiB RAM
- Workload
- Native coding harnesses,
full-stack apps, headless Chromium
What we need to prove
A complete development environment
Package and browser downloads, running apps, authenticated terminals and private previews.
Safe, recoverable execution
Credential isolation, process cancellation and filesystem preservation across sleep and restart.
Economics that work at task scale
Measure startup latency, resource consumption and cost against our current execution backend.
Evaluation intent only. No Daytona integration, partnership, performance or savings claim is implied.
More work done.
Less work babysat.
Building the cloud workspace that lets native coding agents keep going—and brings the work back to you.
Visit statefold.sh ↗