Products · Sail · Environments for AI coding agents
Sail. The layer around your coding agents.
Sail turns bare-metal servers into isolated, spec-driven dev environments where AI coding agents do the work. Each project gets a hard-isolated container. Agents are dispatched against specs with rollback safety and cross-agent review.
- Native binary
- No runtime dependencies
- Sub-millisecond startup
- Linux host · macOS client
- claude-code · codex
- MIT
Quickstart
Running in minutes. Self-serve.
> Install in one line
curl -fsSL https://raw.githubusercontent.com/standardapplied/sail/main/install.sh | bash > Stand up a box and dispatch a spec
sudo sail init --as-main # provision the box and its control plane
sail project init # author a sail.yaml
sail project apply web # create, start, or converge the container
sail spec create --project web --title "Stripe webhook"
sail spec dispatch --project web # snapshot, branch, launch the agent How it works
Simple. Explicit. No magic.
One container per project.
Each project gets its own Incus container with its own filesystem, network, and rootless Podman services. One runaway agent can't touch another project.
One sail.yaml. One verb.
Runtimes, services, repos, and agent config in one file. sail project apply creates, starts, or converges the container from any starting state.
Specs are the unit of work.
Status, assignee, and dependencies live in a database every engineer's box shares. Dispatch picks the next ready spec and launches the agent with full context.
Snapshot, then run.
Every dispatch snapshots the container first. A runaway agent is stopped at its time limit and rolled back to the pre-launch snapshot.
One agent codes. Another reviews.
A review pipeline checks the branch, a fix agent addresses findings, and the loop repeats until it passes or escalates to a human.
One main, many nodes.
Specs and projects sync between engineers' boxes over a locked-down SSH gateway, with three-way conflict resolution so no one's work is overwritten.
Choosing a tool
When Sail fits. And when another tool fits better.
You run coding agents on hardware you own and want one hard-isolated environment per project, spec-driven dispatch, snapshot rollback, and cross-agent review.
You need a reproducible container per repository, opened from your editor, without agent orchestration.
You want managed cloud development environments for a large team, on Kubernetes or VMs.
You need hosted sandboxes that execute agent-generated code on demand through an API.
In the field
Sail provisions every SAIL development environment. It's how a strike team of two or three runs several engagements at once without one agent touching another client's code.
How we work