# Let'em Cook

> Unhinged, Accelerated DevOps for Agents & Users — LEMC turns versioned recipes into discoverable, bounded actions that people click and agents run with lemcli.

## Key pages

- [Home](https://letemcook.dev/) - Product overview, sheriff field notes, and the human/agent action model.
- [Documentation](https://letemcook.dev/docs/) - Clickable HTML documentation presented in a Markdown-shaped interface.
- [Releases](https://letemcook.dev/releases/) - First-party amd64/arm64 lemcli binaries and digest-pinned AMD64 LEMC web and runner images, with SHA-256 and OCI digest proof.

## 1. Start here

- [What is Let'em Cook?](https://letemcook.dev/docs/what-is-lemc/) - The product model: cookbooks, pages, recipes, fresh jailed Firecracker jobs, and durable output.
- [Publish a cookbook and grant the team](https://letemcook.dev/docs/publish-cookbooks/) - Maintained source examples plus the role-by-role workflow for managed container publishing, separate server-side secret bindings, app installation, and run-only user ACLs.
- [Operational use cases](https://letemcook.dev/docs/use-cases/) - BYOK WordPress migrations, corporate Nmap reports, GCP image lifecycle actions, Terraform labs, restore drills, job timing, and governed callback transitions.
- [Deterministic computational units](https://letemcook.dev/docs/deterministic-units/) - What the deterministic boundary guarantees, what it does not, and why it helps agents.

## 2. Humans and agents

- [MCP integration vs LEMC's operating model](https://letemcook.dev/docs/humans-and-agents/) - MCP standardizes AI integration; LEMC packages, authorizes, executes, distributes, and proves reviewed computation across supply chain, authority, isolation, usability, and durable evidence.
- [lemcli field guide](https://letemcook.dev/docs/lemcli/) - Discover actions, submit recipes, follow output, inspect status, and retrieve logs or artifacts.

## 3. Runtime and output

- [Themes: install, select, and govern](https://letemcook.dev/docs/themes/) - Server operators install persistent theme bundles; account admins or account-admin agents select an installed theme. Current releases do not expose a user or agent upload flow.
- [Architecture](https://letemcook.dev/docs/architecture/) - The fresh-per-job Firecracker boundary, digest-pinned runtime, event fanout, teardown, and durable storage flow.
- [Recipe state and LEMC verbs](https://letemcook.dev/docs/lemc-verbs/) - How fresh tasks hydrate durable workspaces, receive context, carry step values, publish artifacts, and emit ordered browser and CLI output.

## Agent quick path

1. Install the verified architecture-specific CLI with `curl -fsSL https://letemcook.dev/lemcli/install.sh | bash`. The installer verifies the selected artifact and installs it as `/usr/local/bin/lemcli` on `PATH`.
2. Run `lemcli init` to connect and authenticate with the target LEMC server.
3. List the apps currently authorized for this identity with `lemcli apps list --format=json`.
4. Inspect a selected action surface with `lemcli jobs surface ... --format=json`.
5. Choose a returned action ID and supply only its declared form values.
6. Submit with `lemcli jobs run ... --action <action-id> --follow --render json`.
7. Confirm task ID, terminal status and reason, system and recipe steps, logs, and artifacts.

In a Ralph-style iterative loop, repeat discovery at the start of each pass so newly distributed, currently authorized tooling can be considered. Never run a new app merely because it appeared; inspect its action contract and use it only when the task requires it.

Execution mode, runner selection, and isolation policy are server/operator decisions. Clients do not choose them per request.