Quickstart

By the end of this page, you'll have a chatbot running on your machine that you can talk to via curl, the Vystak chat client, or any OpenAI-compatible client.

This takes about five minutes.

Prerequisites

• Vystak installed
• Docker running
• An Anthropic-compatible API key (we use MiniMax in this example because they offer a generous free tier)

Step 1: Create a project

mkdir my-first-agent
cd my-first-agent

Step 2: Define the agent

Create a file named vystak.yaml with this content:

name: hello-bot
instructions: |
  You are a friendly assistant. Be concise.
model:
  name: minimax
  provider:
    name: anthropic
    type: anthropic
  model_name: MiniMax-M2.7
  parameters:
    temperature: 0.7
    anthropic_api_url: https://api.minimax.io/anthropic
platform:
  name: docker
  type: docker
  provider:
    name: docker
    type: docker
channels:
  - name: api
    type: api
secrets:
  - name: ANTHROPIC_API_KEY
port: 8090

A quick tour:

• name — your agent's name; also the container name prefix
• instructions — the system prompt
• model — the LLM to call. We're using MiniMax through their Anthropic-compatible endpoint
• platform — where to run. We're targeting Docker
• channels — how users reach the agent. api exposes a REST endpoint
• secrets — env vars the agent needs at runtime
• port — host port to bind on (optional; Docker picks one if omitted)

Step 3: Set your API key

export ANTHROPIC_API_KEY=your-key-here

Step 4: Deploy

vystak apply

You'll see output like:

Loaded 1 agent(s)

Agent: hello-bot
  Validating... OK
  Generating code... OK
  Deploying:
    Building Docker image... OK
    Starting container... OK

============================================================
Deployment complete — 1 agent(s) deployed
============================================================

Shared Infrastructure:
  Provider:     Docker (local)
  Network:      vystak-net

Agents:
  hello-bot    http://localhost:8090

Connect:
  vystak-chat --url http://localhost:8090

Vystak just generated a LangGraph agent + FastAPI server, packaged them in a Docker image, and started a container.

Step 5: Talk to it

The simplest way is the interactive chat client:

vystak-chat --url http://localhost:8090

Or one-shot:

vystak-chat --url http://localhost:8090 -p "Hello!"

Or via curl:

curl -X POST http://localhost:8090/invoke \
  -H "Content-Type: application/json" \
  -d '{"message": "Hello!"}'

Or via any OpenAI-compatible client — the agent serves /v1/chat/completions:

curl http://localhost:8090/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "hello-bot",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Step 6: Iterate

Edit vystak.yaml — change the instructions, add temperature: 0.5, whatever — then run vystak apply again. Vystak hashes the definition and only redeploys if something actually changed.

Step 7: Tear down

vystak destroy

This stops and removes the container. The Docker network and any volumes stick around so a redeploy is fast.

What you just did

1. Defined an agent in 17 lines of YAML.
2. Generated a LangGraph agent + FastAPI harness from that definition.
3. Built and deployed the agent as a Docker container.
4. Talked to it via three different clients.

Vystak handled the LangChain integration, the Dockerfile, the networking, and the lifecycle. You wrote infrastructure as code.

What's next

• Agents — add tools, sessions, memory, multiple channels
• Examples — more complex setups including multi-agent
• Deploying to Docker — how the Docker provider works