What is Llanite?

What Llanite installs

When you run llanite install, Llanite configures the model, the local runtime that serves it, the agent layer that controls the prompt loop, the tools the agent can call, the permission policy for those tools, and the hardware fit metadata for your machine. Every component is recorded in an inspectable stack manifest. Nothing about the setup is hidden — you can read the full stack configuration before anything runs with llanite inspect.

$ npm install -g @llanite/cli$ llanite doctor

Who Llanite is for

Llanite is for developers who want to run AI agents locally without sending prompts to a cloud API. That includes people building private coding workflows, teams who need reproducible AI setups across different machines, and developers who want to understand exactly what tools and permissions an AI agent has before running it. If you have tried wiring together Ollama, an agent tool, and a set of permissions manually, Llanite is the layer that keeps those pieces together and inspectable.

Why stacks matter

Local AI setups fail most often because the model, runtime, agent, and permission settings were never designed to work together as a unit. A model that is good at conversation may produce poor tool calls. A context window that is too large slows every response. An agent with unrestricted filesystem access is a security risk. Llanite stacks encode all of these choices together so you can inspect them before anything runs, share them with teammates, and reproduce the same setup in a new project without rediscovering the configuration.

Core commands

The core Llanite workflow is four commands: install the CLI, check prerequisites, browse available stacks, and run one. Once a stack is running, you interact with the agent in a terminal UI that shows the model, runtime, active permissions, and live RAM usage.

$ llanite stacks$ llanite inspect standard$ llanite install standard$ llanite run standard

How Llanite compares to running Ollama alone

Ollama handles local model serving — it downloads models and runs the inference server. Llanite adds everything that turns a served model into a useful agent: the prompt loop, the tool surface, the permission policy, the hardware fit check, and a terminal UI for the session. Running Ollama alone gives you a local model you can query. Running a Llanite stack gives you a complete local agent workflow where the model, agent, and tool access are configured and reproducible out of the box.

Getting started

Llanite is distributed via npm and requires Node 20 or later. You also need Ollama installed and running — llanite doctor will tell you if it is not reachable. Once prerequisites pass, browse available stacks and pick one that fits your machine's RAM.

1. 1Install Llanite: npm install -g @llanite/cli
2. 2Start Ollama: ollama serve
3. 3Check prerequisites: llanite doctor
4. 4Browse stacks: llanite stacks
5. 5Install and run a stack: llanite install standard && llanite run standard