Skip to content

Getting Started

Jump to bottom
lacausecrypto edited this page Apr 6, 2026 · 2 revisions

Getting Started

Get OCC running and execute your first chain in under 5 minutes.

Prerequisites

  • Node.js >= 20 (download)
  • Claude CLI installed and authenticated: 
    npm install -g @anthropic-ai/claude-code
claude auth login

Note: Claude CLI is required for the default claude provider. You can also use OpenRouter, OpenAI, Ollama, or HuggingFace without Claude CLI.

Installation

git clone https://github.com/lacausecrypto/OCC.git
cd OCC/mcp-server
npm install
npm run build

Configuration

Option A: REST API (standalone)

No config needed — just run:

npm run rest
# Server running on http://localhost:4242

Option B: With React Dashboard

# Terminal 1: Backend
cd OCC/mcp-server && npm run rest

# Terminal 2: Frontend
cd OCC/frontend-react && npm install && npm run dev
# Dashboard at http://localhost:5173

Option C: MCP (for Claude Code / Claude Desktop)

  1. Copy the example config:

    cd ..  # back to OCC root
cp .mcp.json.example .mcp.json

  2. Edit .mcp.json with your absolute paths:

    {
  "mcpServers": {
    "chain-orchestrator": {
      "command": "node",
      "args": ["/absolute/path/to/OCC/mcp-server/dist/index.js"],
      "env": {
        "CHAINS_DIR": "/absolute/path/to/OCC/chains",
        "PIPELINES_DIR": "/absolute/path/to/OCC/pipelines",
        "REST_PORT": "4242"
      }
    }
  }
}

  3. Start:

    cd mcp-server
npm start

Option D: Docker

docker compose up -d
# Backend: http://localhost:4242
# Frontend: http://localhost:5173

Run Your First Chain

Via REST API

# Execute the deep-researcher chain
curl -X POST http://localhost:4242/execute/deep-researcher \
  -H "Content-Type: application/json" \
  -d '{"input": {"topic": "quantum computing breakthroughs"}}'

# Response: {"executionId": "1a2b3c4d..."}

# Check status
curl http://localhost:4242/executions/1a2b3c4d

# Stream real-time progress
curl -N http://localhost:4242/executions/1a2b3c4d/stream

Via React Dashboard

  1. Open http://localhost:5173
  2. Click on a chain in the sidebar
  3. Fill in inputs and click Run
  4. Watch real-time SSE mini-terminals per step

Via MCP (in Claude Code)

Simply ask Claude:

Run the deep-researcher chain on "quantum computing breakthroughs"

Claude will use the run_chain MCP tool automatically.

Add LLM Providers

Go to Settings > LLM Providers in the dashboard to add:

  • OpenRouter — 200+ models (GPT-4o, Gemini, Llama, DeepSeek...)
  • OpenAI — GPT-4o, o3-mini
  • Ollama — Local models (Settings > Ollama to pull models)
  • HuggingFace — 118 free models via Inference API (Settings > HuggingFace)

Each step in a chain can use any model from any enabled provider via the model field.

Create Your First Chain

Create chains/hello-world.yaml:

name: hello-world
description: "My first OCC chain"
version: "1.0"

inputs:
  - name: name
    description: "Who to greet"

steps:
  - id: greet
    prompt: "Write a creative greeting for {input.name}. Be fun and original."
    output_var: greeting

  - id: poem
    depends_on: [greet]
    prompt: |
      Based on this greeting: {greeting}

      Now write a short 4-line poem for {input.name}.
    output_var: poem

output: poem

Execute it:

curl -X POST http://localhost:4242/execute/hello-world \
  -H "Content-Type: application/json" \
  -d '{"input": {"name": "Alice"}}'

What's Next?

Clone this wiki locally