Solutions · Coding agents

Coding agents need somewhere to work.

The agent loop is excellent. What it doesn’t come with is a place to run code, files that survive a handoff, a database to read, or a trace of what it did. sys9 is that substrate — atomic services you compose under Claude Code, Codex, OpenClaw, Cursor, or OpenCode with the setup required by each runtime.

Get started See run9 →
Keep your runtimeExplicit credentialsSeparate task environments
what you compose

A workspace, assembled from four services.

Each does one job. Start with a sandbox; add files, a database, and traces as the agent needs them.

run9

A sandbox per task

Fork a Root Snap containing the OS, repository, and installed packages, then start a separate Box for a task. Persistent Volumes are separate; running processes and credentials need their own setup.

run9 → drive9

Files that follow the agent

A shared FUSE filesystem across sandboxes, sessions, and handoffs — plus semantic grep (vector + BM25 in parallel) so the agent finds the right file, not just the named one.

drive9 → db9

A database from the terminal

Serverless Postgres with database branching and built-in file storage. Give a review its own database and scoped access.

db9 → owl9

Capture agent sessions

Install owl9, configure a private space and the runtime integration, then enable capture for supported sessions. Inspect native transcripts and linked sub-sessions.

owl9 →
the shape

Create a task environment, run the agent, keep the transcript.

Start each task from a prepared Root Snap. Run the agent inside the new Box, export the result, and remove resources you no longer need. Host Compute and retained storage have separate usage.

one task, one sandbox

A separate workspace for each task.

First install the sys9 CLI and complete run9 login. Prepare a repo-template Box with the repository at /work/repo and Claude Code plus owl9 installed. Keep provider credentials out of the reusable template. After creating the task Box, configure its provider, a unique private owl9 space, and runtime integration before enabling capture. Replace the Snap placeholder with the ID returned by the fork command; choose a unique Box name for each task.

Root Snap fork project quotas apply configured session capture SSH-drivable
terminal
# Stop the prepared template, then fork its Root Snap
sys9 run box stop repo-template
sys9 run snap fork --from-box repo-template
sys9 run box create task-41 --snap '<returned-root-snap-id>'

# Configure provider and owl9 inside task-41 first
sys9 run box exec task-41 /bin/sh -lc \
  'cd /work/repo && OWL9_ENABLE=1 claude -p "fix the failing test"'

# Export the result before removing the task Box
sys9 run box cp task-41:/work/repo ./task-41-result
sys9 run box rm task-41
works with your loop

Bring the agent. We bring the substrate.

Use the services with your existing runtime. Session capture requires a supported owl9 integration; follow its setup instructions in the environment where the agent runs.

Claude CodeOpenAI CodexOpenClawCursorOpenCodeCline
related products

The services behind this shape.

run9

Forkable sandbox

Fork a filesystem Snap, create a Box, and execute through the CLI or SSH.

run9 → drive9

One filesystem

Shared FUSE filesystem, semantic grep, per-agent secrets vault.

drive9 → db9

Postgres, from terminal

Serverless Postgres, database branching, and built-in file storage.

db9 → owl9

Session observability

Capture supported sessions after setup; read transcripts and linked sub-sessions.

owl9 →
start with a sandbox

Give your coding agent a real workspace.

Install the CLI, sign in to run9, and follow the quickstart to create your first Box.

Open run9 quickstart Running several at once? →