Getting started

This guide walks through the full path from zero to a working inference endpoint:

Register a machine (install mantlerd)
Build a stack in the Forge
Deploy the stack to your machine
Call the endpoint

1. Register a machine

mantlerd is the daemon that runs on each worker machine. Install it using the release script.

Linux (recommended — run as root):

curl -sSL https://raw.githubusercontent.com/Borgels/mantlerd/master/scripts/install.sh | \
  sudo sh -s -- \
  --token YOUR_MACHINE_TOKEN \
  --machine MACHINE_ID \
  --server https://control.mantler.ai

macOS:

curl -sSL https://raw.githubusercontent.com/Borgels/mantlerd/master/scripts/install.sh | \
  sh -s -- \
  --token YOUR_MACHINE_TOKEN \
  --machine MACHINE_ID \
  --server https://control.mantler.ai

Your machine token and ID come from the Machines page in the Mantler web app — add a machine there to generate them.

After installation, verify the daemon is running:

mantler doctor

The machine will appear as online in the web app within a few seconds.

2. Build a stack in the Forge

Open the Forge in the Mantler web app. A stack (mantle) is built from layers:

Layer	Role	Required
Machine	The machine to deploy to	Yes
Runtime	Inference runtime (Ollama, vLLM, llama.cpp, MLX, …)	Yes
Model	The language model	Yes
Harness	Prompt routing / tool wrapper	Optional
Orchestrator	Multi-step orchestrator	Optional
Endpoint	OpenAI-compatible endpoint exposure	Optional, but needed for API access

Select your machine from the Machine slot, pick a runtime, pick a model. The compatibility panel shows in real time whether the combination will work on your hardware.

Add an Endpoint layer to expose the stack as an API endpoint.

Click Forge to create the mantle.

3. Deploy the stack

From the mantle detail page, click Deploy. The Mantler control plane sends a sequence of commands to mantlerd on your machine: install the runtime, pull the model, configure the endpoint.

You can monitor progress on the machine detail page or with:

mantler info
mantler runtime list
mantler model list

4. Make your first API call

Once deployed, the mantle's Endpoint layer is active. From your API keys page, create an mk_live_ key.

Test with curl:

curl https://api.mantler.ai/v1/chat/completions \
  -H "Authorization: Bearer mk_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_MODEL_ID",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

To use in Cursor or VS Code extensions, set the OpenAI base URL to https://api.mantler.ai/v1 and use your API key.

What's next

Machine configuration — tune intervals, trust mode, allowed commands
CLI reference — all mantler subcommands
API reference — full request/response spec