This guide runs the full ManyLayers stack on your machine from a source checkout, then takes you to a first chat completion. You need Docker, make, and an API key for one AI provider (for example OpenAI).
1

Start the stack

make env-files     # create infra/docker/.env.dev from its template
make dev-migrate   # apply database migrations (no `up` target migrates)
make dev-up        # postgres, redis, gateway, workspace, deployer
This builds the images from source and starts three services:
ServiceURLWhat it serves
Gatewayhttp://localhost:8180The OpenAI-compatible /v1 API
Workspacehttp://localhost:8190/ui/The console and the admin API
Deployerhttp://localhost:8200Model deployments and training jobs
Postgres is published on localhost:5532 and Redis on localhost:6479.
2

Verify the gateway is healthy

curl http://localhost:8180/healthz

{"status":"ok"}
/readyz additionally checks Postgres and Redis.
3

Create your organization

Open http://localhost:8190/ui/ and choose Sign up. Enter an organization name, your email and a password (at least 8 characters). The organization is active immediately and you are its owner.
4

Connect a provider

In the console, open Providers, add a provider account (for example type openai), paste its API key, and enable the models you want to serve, such as gpt-4o-mini.Console providers serve /v1/chat/completions, /v1/responses and /v1/messages. For embeddings, images, audio or rerank, or for Azure, Bedrock and Vertex, declare the model in gateway.yaml instead. See Providers.
5

Create a personal access token

Open API Keys and create a personal access token for your workspace. Copy the ml_pat_... value; it is shown only once.
6

List available models

curl http://localhost:8180/v1/models \
  -H "Authorization: Bearer ml_pat_..."
You’ll see the models you enabled.
7

Make your first chat completion

curl http://localhost:8180/v1/chat/completions \
  -H "Authorization: Bearer ml_pat_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini",
    "messages": [
      {"role": "user", "content": "Hello, ManyLayers!"}
    ]
  }'
8

Explore the console

Back at http://localhost:8190/ui/ you can:
  • Open the request you just made in Request Traces
  • Create teams, API keys and virtual accounts
  • Configure routing, rate limits, budgets and guardrails
  • Use chat, knowledge bases and agents in the workspace
For the full stack with the optional backends (Kafka, Qdrant, MinIO, dstack), run make up instead of make dev-up. make down stops everything.

​
Next steps

AI Gateway quick start

Connect providers via the admin API or gateway.yaml, and issue API keys.

Connect a real provider

Point the gateway at any of the 20 supported provider types.

Gateway configuration

Walk through every section of the gateway YAML config.

Self-host the gateway

Run ManyLayers in production.