Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/archestra-ai/archestra/llms.txt

Use this file to discover all available pages before exploring further.

Getting Archestra running locally takes a single docker run command. The all-in-one Docker image bundles the Admin UI, the API, a PostgreSQL database, and a KinD (Kubernetes-in-Docker) cluster for MCP server orchestration — so you have the full platform available on your machine without any external dependencies. Once it’s up, you’ll build a real agent, point it at a live website, and then connect it to Claude as an MCP server to experience the full end-to-end flow.

​
Step 1 — Pull and Run the Platform

Copy the command for your operating system and run it in a terminal. Both variants expose the Admin UI on port 3000 and the API on port 9000. 
docker pull archestra/platform:latest;
docker run -p 9000:9000 -p 3000:3000 \
   -e ARCHESTRA_QUICKSTART=true \
   -v /var/run/docker.sock:/var/run/docker.sock \
   -v archestra-postgres-data:/var/lib/postgresql/data \
   -v archestra-app-data:/app/data \
   archestra/platform;
Once the container is running, open http://localhost:3000 in your browser.
The -v /var/run/docker.sock:/var/run/docker.sock mount enables the embedded KinD Kubernetes cluster used to run MCP servers. It is required for the quickstart Docker deployment. For production, use the Helm deployment with an external Kubernetes cluster instead. If you’re accessing Archestra from another device on your network, private network IPs (e.g. 192.168.x.x, 10.x.x.x) are automatically trusted in quickstart mode.

​
Step 2 — Build Your First Agent

This walkthrough creates an agent that can read and answer questions about the Archestra documentation using a headless Playwright browser — no coding required.
1

Install the Playwright MCP Server

Go to MCP Registry, search for microsoft__playwright-mcp, and install it. Watch Archestra spin up the MCP server as an isolated pod in the embedded Kubernetes cluster. If you want to confirm it’s running: kubectl get pods
2

Create the Agent

Go to Agents and create a new agent named “Archestra Docs Reader Agent” with the following system prompt:
You're using playwright to answer questions about Archestra based on https://archestra.ai/docs/
3

Enable MCP Tools for the Agent

While editing the agent, enable all microsoft__playwright-mcp tools so the agent can control the browser.
4

Add an LLM API Key

Go to Settings → LLM API Keys and add a key for your preferred provider. You can connect commercial providers like OpenAI, Anthropic, or Google Gemini. For a free option, use Cerebras or a local Ollama instance.
5

Chat with Your Agent

Go to Chat, choose the “Archestra Docs Reader Agent” from the agent selector, and ask:
How could I deploy Archestra?
Watch the agent navigate the Archestra docs website in real time to answer your question.

​
Step 3 — Connect to Your Agent via MCP Gateway

Archestra is not just a Chat UI — it’s a fully capable MCP Gateway. This advanced step connects your “Archestra Docs Reader Agent” to Claude Code so it can be invoked as an MCP server from any MCP-compatible client. The request flow looks like this: 
Claude → Archestra MCP Gateway → MCP Orchestrator (K8s) → Playwright Headless Browser
1

Create an MCP Gateway

Go to MCP Gateways and create a new gateway. Set “Archestra Docs Reader Agent” as a sub-agent, then save and copy the generated MCP configuration.
2

Add the Gateway to Claude

The copied configuration will look similar to this (your URL and auth key will be different):
{
  "mcpServers": {
    "archestra": {
      "url": "http://localhost:9000/v1/mcp/e1b0272c-3839-4575-a49d-aabb864d638d",
      "headers": {
        "Authorization": "Bearer arch_119220a7bfc485d66b678d3e9fb2db36"
      }
    }
  }
}
You can also add it via the Claude CLI:
claude mcp add archestra "http://localhost:9000/v1/mcp/e1b0272c-3839-4575-a49d-aabb864d638d" \
  --transport http \
  --header "Authorization: Bearer arch_119220a7bfc485d66b678d3e9fb2db36"
3

Ask Claude to Use the Gateway

In Claude, prompt it to use the Archestra MCP server:
Ask archestra to give you all the deployment options.
Claude will call your Archestra MCP Gateway, which will invoke the Playwright agent inside Kubernetes, navigate the docs, and return a structured response. ✨

​
What’s Next

Chat Interface

Learn how to interact with AI agents using MCP tools through the Agentic Chat UI.

Agents

Build autonomous agents with no code — system prompts, tools, triggers, and sub-agents.

LLM Proxy

Secure your AI applications with the drop-in LLM proxy and virtual API keys.

MCP Gateway

Connect external systems and clients to your agents through a single authenticated endpoint.

Deployment

Deploy Archestra to production with Helm, external PostgreSQL, and cloud provider configuration.

Join the Community

Connect with other users and the Archestra team in the Slack community.

Build docs developers (and LLMs) love

Get started for freeTalk to us