pilot run / isolated-compute

> Generated code should run somewhere disposable.
Not on the laptop that hired it.

Give coding agents a real execution environment with a visible lifecycle: verify the tool, launch a microVM or container, attach only the data it needs, and return an artifact instead of host access.

For coding-agent and platform teams that need useful execution without handing the agent an unrestricted host shell.
job/8f42c1isolated

01$ pilot app call smol.run

02creating microVM........ done

03mount /workspace.......... read-only

04network egress........... denied

05execute analysis.py...... running

RESULT3 files · 1 chart · exit 0cpu 1.2s · memory 84MB · artifacts retained 24h
Boundary: hardware-isolated microVMAudit: enabled
Containment model

Disposable by design.
Useful by contract.

Coding agents need to run commands, inspect artifacts, and persist working data. Running everything directly on a developer machine combines the broadest authority with the least repeatability. Pilot separates the agent call from the execution substrate, allowing teams to choose microVM isolation, a container, or a purpose-built database app for the task.

Host
Supervisor
microVM
generated jobbounded input / output
01Isolate

Hardware-isolated microVMs

Smol runs workloads in local or cloud microVMs with an explicit lifecycle and brokered control surface.

02Package

Container execution

The Docker app provides a familiar unit for repeatable tools, dependencies, and disposable environments.

03Persist

Typed local data services

SQLite, DuckDB, PostgreSQL, MySQL, and Redis cover transactional, analytical, and in-memory workloads.

Job lifecycle

The host receives artifacts, not assumptions.

  1. 01

    Choose the isolation level

    Match a microVM, container, or data process to the workload and threat model.

  2. 02

    Declare the required authority

    App grants expose the files, process calls, and network access the tool requires.

  3. 03

    Run through typed methods

    The agent invokes a bounded operation instead of receiving a permanent unrestricted shell.

  4. 04

    Destroy or retain deliberately

    Ephemeral compute and persistent data have separate, visible lifecycles.

Sandbox selector

Choose the right boundary
for generated code.

Describe the workload, its data, and any network requirement. We will recommend a microVM, container, or narrower data-app execution path.

What you get
  1. 01Isolation substrate
  2. 02Resource and network limits
  3. 03Artifact boundary
Start with one real workflowAbout 2 minutes

Do not include credentials, secrets, or customer data. By submitting, you agree that Pilot may use this information to respond. Privacy Policy.

Technical basis

Specific product surfaces.
Inspectable in the docs.

Product reviewPilot release 1.13.9Inspect documentation
  • 01Local and cloud microVM execution
  • 02Docker lifecycle methods
  • 03Five local database engines
  • 04Signed apps with explicit grants