Get your first agent live on Sublayr in under ten minutes. This guide walks through installing the SDK, connecting a model, configuring a minimal agent, and deploying it to production — with nothing extra to set up first.

Before you start

You'll need a Sublayr account and an API key from your dashboard under Settings > API Keys. You'll also need at least one model provider key (OpenAI, Anthropic, Mistral, or your own endpoint) to connect during setup.

Install the SDK

Install the CLI and SDK via npm:

npm i @sublayr/sdk

npm i @sublayr/sdk

npm i @sublayr/sdk

Verify the install:

sublayr --version

sublayr --version

sublayr --version

Authenticate

Log in with your API key:

sublayr auth login

sublayr auth login

sublayr auth login

This stores your credentials locally so subsequent commands don't require re-authentication.

Connect a model

Connect the model provider you want your agent to run on:

sublayr models connect anthropic

sublayr models connect anthropic

sublayr models connect anthropic

You can connect multiple providers and choose one per agent, or set a fallback model for automatic rerouting if your primary provider degrades.

Write a minimal config

Create an agent.yaml file:

name: support-agent
model: anthropic/claude
fallback: openai/gpt-4o
tools:
  - postgres
  - slack
budget: 200/mo

name: support-agent
model: anthropic/claude
fallback: openai/gpt-4o
tools:
  - postgres
  - slack
budget: 200/mo

name: support-agent
model: anthropic/claude
fallback: openai/gpt-4o
tools:
  - postgres
  - slack
budget: 200/mo

This defines an agent named support-agent, running on Claude with GPT-4o as a fallback, connected to Postgres and Slack, capped at $200/month.

Deploy

Push it live:

sublayr deploy --config agent.yaml

sublayr deploy --config agent.yaml

sublayr deploy --config agent.yaml

Sublayr validates the config, provisions the runtime, and returns an agent ID and live endpoint.

Verify it's running

sublayr agents status agt_4b81

sublayr agents status agt_4b81

sublayr agents status agt_4b81

This returns the agent's current status, uptime, and the last few runs. From here, every run is automatically traced — head to Observability in your dashboard to watch it work in real time.

Next steps

• Set up approval rules for high-stakes actions

• Configure budgets and spend alerts

• Explore the full API reference for programmatic control