Skip to content
rossoctlPublic

About

Run stateful AI coding agents on a dense, warm process-manager runtime — sessions resume exactly where they left off.

Resources

Code of conduct

Contributing

Security policy

Stars

12 stars

Watchers

0 watching

Forks

Latest commit

 

History

3,855 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

moca

Run AI coding agents' work away from where they run: in a sandbox service for the agent on your laptop, or on a harness that runs the agent sessions themselves.

status node license

Moca has two capabilities. Use either one on its own, or both together:

  • Sandbox service. Your coding agent (Claude Code, OpenCode, Codex or IBM Bob) keeps running on your laptop. Every command it runs goes to a fresh, isolated VM on a shared service, working on a copy of your repository. You take the agent's work back as a patch. Your laptop's source files are untouched until you apply it, apart from the agent configuration mocactl sbx up writes.
  • Harness. Moca runs the agent itself. A supervisor hands each connection to a pool of warm workers, sessions are durable in Redis and resume by id, and tools run in a remote sandbox (a container, or a Firecracker microVM per command). An orchestrator dispatches work over HTTP.

📖 Documentation: rossoctl.dev/moca, built from docs/.

Try it

Harness, on your laptop, in about five minutes. You need Docker with Compose, curl, and a model key:

curl -fsSL https://raw.githubusercontent.com/rossoctl/moca/main/deploy/compose/install.sh \
  | ANTHROPIC_API_KEY=sk-ant-... sh

curl -s -H 'Content-Type: application/json' -d '{"prompt":"Run uname -a and tell me the kernel."}' \
  http://127.0.0.1:8080/turn

Next step: the harness quickstart.

Sandbox service. You need a sandbox service deployment, its URL, and a GitHub account enrolled in it:

curl -fsSL https://raw.githubusercontent.com/rossoctl/moca/main/scripts/install-mocactl.sh | sh
mocactl sbx login --control-plane-url https://<api host>
cd <your repository> && mocactl sbx up     # then run claude or opencode as usual
mocactl sbx pull --apply                   # take the agent's work back

Next step: the sandbox quickstart.

Documentation

You are… Start at
Evaluating moca Overview: the two capabilities, how they relate, and when to use each
Using the sandbox service Sandbox quickstart
Building on the harness Harness quickstart
Deploying or operating moca Deploy and operate: which deployment delivers which capability

Status

Every deployment path is Preview: this is explorative work, and interfaces may change. See Maturity and lifecycle.

Path Delivers Label
Docker Compose Harness (trial) Preview, trial only
Single VM Harness Preview
Kubernetes / OpenShift Harness Preview, first target for Stable
MicroVM tier (P4) The harness's microVM sandboxes Preview
Sandbox service Sandbox service Preview

Roadmap and the evidence behind moca's design: Project (a draft).

Releases: GitHub releases. Images: ghcr.io/rossoctl/moca and ghcr.io/rossoctl/moca-remote-worker.

Repository layout

Path What it is
harness/, packages/server, packages/supervisor, packages/session-backend The harness runtime: Pi agent adapter, worker and server entrypoints, the P6 supervisor, Redis sessions
packages/control-plane Login, sessions, credentials and sandboxes (route profiles harness, sandbox, combined)
packages/sandbox-relay, packages/k8s-sandbox, packages/sandbox-client The gRPC relay, sandbox transports, and the sandbox/v1 client with its MCP server
remote-worker/ Go: container and microVM sandbox workers, egress proxy, vmpoolctl
packages/mocactl mocactl, the client for both capabilities
deploy/ compose/, vm/, k8s/, microvm/, sbx/ deployments
docs/ The documentation source, published at rossoctl.dev/moca

Contributing

See CONTRIBUTING.md and Contributing to the docs. DCO sign-off is required (git commit -s).

License

Apache-2.0

About

Run stateful AI coding agents on a dense, warm process-manager runtime — sessions resume exactly where they left off.

Resources

Code of conduct

Contributing

Security policy

Stars

12 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages