Llanite guide
Llanite vs Ollama
Ollama and Llanite address different parts of the local AI stack. Ollama is a local model runtime — it downloads models and serves them through a local API. Llanite is a stack manager — it packages the full workflow around a model, including the agent loop, tool access, permission policy, and hardware requirements. The two tools are designed to work together, not to replace each other.
What Ollama does
Ollama is a tool for running large language models locally. It downloads model weights, manages them on disk, and runs a local HTTP server that exposes a model API compatible with OpenAI-style clients. Ollama handles hardware acceleration — Metal on Apple Silicon, CUDA on Nvidia — which means you get reasonably fast local inference without configuring drivers manually. Ollama is the most widely used local model runtime and is the runtime that Llanite currently targets.
What Llanite adds
Llanite sits on top of a local runtime like Ollama and adds the agentic layer. This includes the prompt loop that decides when to call tools, the tool definitions themselves (filesystem, shell, git, web search), the permission policy that controls what each tool is allowed to do, and the terminal UI that shows the session state, memory usage, and live metrics. Llanite also provides a registry of pre-configured stacks that pair models with agent configurations known to work well together.
How they work together
A typical Llanite setup uses Ollama as its runtime. When you run llanite install, Llanite resolves the model declared in the stack manifest and pulls it via Ollama if it is not already cached. When you run llanite run, the agent session starts and communicates with Ollama's local API to send and receive model completions. Ollama handles model serving; Llanite handles everything that happens around it.
$ # Start Ollama$ ollama serve$ $ # Install and run a Llanite stack$ llanite install local-coder$ llanite run local-coderWhen to use each
Use Ollama alone when you want to query a local model directly, experiment with models in a notebook, or build your own tooling around the Ollama API. Use Llanite when you want a complete local agent workflow — model, runtime, agent, tools, and permissions — packaged and inspectable. Most Llanite users already have Ollama installed; Llanite extends what they can do with it rather than replacing it.
Getting started with both
Install Ollama first, then install Llanite via npm. Run llanite doctor to confirm both are working before installing a stack.
$ npm install -g @llanite/cli$ llanite doctor$ llanite install local-coder$ llanite run local-coder- 1Install Ollama from ollama.com
- 2Start Ollama: ollama serve
- 3Install Llanite: npm install -g @llanite/cli
- 4Run llanite doctor to verify the setup
- 5Install a stack: llanite install local-coder
- 6Run the stack: llanite run local-coder
Comparison
| Feature | Ollama | Llanite |
|---|---|---|
| Local model serving | ✓ | Via Ollama runtime |
| Agent loop (ReAct) | ✗ | ✓ |
| Tool permissions | ✗ | ✓ |
| Stack management | ✗ | ✓ |
| Hardware fit check | ✗ | ✓ |
| Terminal UI | ✗ | ✓ |
| Open source | ✓ | ✓ |
| CLI | ✓ | ✓ |
FAQ
Does Llanite replace Ollama?
No. Llanite uses Ollama as its model runtime. Ollama is responsible for downloading and serving models; Llanite adds the agent loop, tool surface, and permission management on top.
Can I use Llanite without Ollama?
Not with the current stacks. Llanite currently targets Ollama as its local runtime. Ollama needs to be installed and running before Llanite can serve a model.
I already use Ollama — do I need Llanite?
Not necessarily. If you are querying models directly or have built your own tooling, Ollama alone is sufficient. Llanite is useful when you want a complete, pre-configured agent workflow with tools, permissions, and a terminal UI without building and maintaining that setup yourself.
Does Llanite work with runtimes other than Ollama?
Llanite's architecture supports swapping runtimes, but currently ships with Ollama as the default and only supported runtime. Support for additional runtimes may be added via the registry.
Is Ollama required before installing Llanite?
You can install the Llanite CLI without Ollama, but you cannot run a stack until Ollama is installed and reachable. Run llanite doctor after installing both to confirm they can communicate.
What order should I set things up?
Install Ollama first, start the Ollama server with ollama serve, then install Llanite with npm install -g @llanite/cli, and finally run llanite doctor to confirm everything is working before installing a stack.