HAMANTOU / PLATFORMThe environment behind
your autonomous workflows.
Explore the product architecture for agent workspaces, isolated execution, and durable task context.
A place for the whole task.
Not just the next prompt.
Our platform roadmap brings together three priorities: persistent workspaces, isolated execution, and stateful workflows.
⌘Agent workspaces
A dedicated environment for project files, dependencies, and intermediate outputs. The workspace is the unit of an agent’s ongoing work.
- Project-scoped files and working directories
- Environment setup and dependency lifecycle
- Artifact collection and handoff
▣Isolated execution
A boundary between your application and the code your agents generate. Execution should be explicit, inspectable, and separate from the host application.
- Command execution and output handling
- Resource limits and task boundaries
- Controlled access to external services
↻Stateful workflows
Longer tasks need checkpoints. A lifecycle built around saving progress and returning to it can support human review and multi-step workflows.
- Checkpoint and restore design
- Pause-and-resume workflow
- Context-aware continuation
Security starts
with clear boundaries.
Our security design priorities address execution boundaries, credential access, data lifecycle, and operational visibility.
01Execution isolation
Evaluate tenant separation, privilege boundaries, and the effects of running untrusted agent-generated code before production use.
02Network & credentials
Define which services a workspace may access, how credentials are delivered, and how access is revoked at the end of a task.
03Data lifecycle
Data lifecycle design covers storage location, retention, deletion, and export requirements for each workload.
04Operational visibility
Define the logs, failure events, resource reporting, and support escalation paths required by your workload.