# Loa Workspace

Limit the reach of AI with isolated environments. Workspace creates sandboxes for agents on hardware you own, or on a VM. An agent gets a set of tools over MCP, scoped to one repository, so it can read, search, edit, run commands, commit and push. Commands stay inside the sandbox, and secrets stay out of the model's reach.

Available on Mac and Linux, and as an experiment in Amazon Bedrock AgentCore.

- Coming soon
- See an example: https://workspace.getloa.app/en/#run

Claude Sonnet, given one task, finds two bugs through the server's tools, fixes them, tests and commits. Nothing here is staged or edited.

- Workspace v0.10.6, the published macOS build
- Claude Sonnet 5.5 in Claude Code
- 8 calls in 15 seconds
- 2026-11-16

The whole run, word for word: https://workspace.getloa.app/en/run/index.md

## The wiring

- Your agent (Claude Code, or any MCP client) → Workspace (tools, confinement, secrets, log): MCP over HTTP, with a bearer token
- Workspace (tools, confinement, secrets, log) → Your repository (the one place commands can see): confined commands
- Workspace (tools, confinement, secrets, log) → One git remote (the only place a push can go): push, with the token
- Workspace (tools, confinement, secrets, log) → Hosts you allow (package registries, docs, APIs): through the proxy

## Inside the box

### Every command is sandboxed

Each command runs in its own sandbox, using bubblewrap on Linux and a Seatbelt profile on macOS. It can write to the repository and a temporary folder, and nowhere else. Each time the server starts, it tries to break out of its own sandbox, and if any attempt works, it won't start. You can turn the sandbox off with sandbox: none, and then the server refuses a git token.

Source, in workspace at v0.11.0: engine/sandbox/bubblewrap.go; engine/sandbox/seatbelt.go; daemon/credential.go; docs/serve.md, "What keeps the daemon from a command", "The proof at every start"

### The git token stays with the server

The token is kept in the macOS keychain, or on Linux in a file only the server's user can read. The server passes it to git only when it pushes to the remote you set up. A push goes to the branch of the same name and is never forced. The model never sees the token, and it's never logged.

Source, in workspace at v0.11.0: docs/serve.md, "The repository's credential"; docs/decisions/git-credential.md §2; docs/getting-started.md §9

### You decide what it can reach

Commands can only reach the hosts you allow, through the server's proxy. On your own machine, the server listens only locally. In Amazon Bedrock AgentCore, AWS checks each caller first. Either way, every request needs a token.

Source, in workspace at v0.11.0: docs/serve.md, "Where it listens", "The token", "Platform agentcore"; daemon/token.go; egress/guard.go; docs/notes/2026-11-20-report-agentcore-experiment.md

### Every call is logged

The log has one line per call, with the tool, how long it took, and whether it was refused. Arguments, headers and tokens are never written to it.

Source, in workspace at v0.11.0: docs/getting-started.md §10; the transcript's record

### Tested on every change

More than 700 automated tests run on every change, including tests that try to escape the sandbox on Linux. The same tests run on a Mac whenever the sandbox code changes. We also test it with coding agents doing real tasks.

Source, in workspace at v0.11.0: .github/workflows/ci.yml, seatbelt.yml; engine/escape_test.go; engine/hostile_test.go; docs/notes/first-sessions-report.md

Loa Workspace v0.11.0. Licensed under the Functional Source License 1.1, with MIT as its future licence (FSL-1.1-MIT): each release becomes MIT two years after it comes out. Coming soon: the download opens when the repository goes public.
