backboard

Files (1)

• SKILL.md

SKILL.md (first lines)

# Backboard

Integration with the [Backboard.io](https://backboard.io) platform for model
routing, memory bridging, and seamless coexistence between R-CLI and ox
agent workflows.

## Installation

### Prerequisites

- ox platform installed
- Backboard.io API key ([dashboard](https://backboard.io/app))
- Node.js >= 18

### Step 1 — Install Backboard MCP adapter

```bash
npm install -g @backboard/mcp-adapter
```

### Step 2 — Configure with ox

Add to `~/.config/ox/mcp.json`:

```json
{
  "mcpServers": {
    "backboard": {
      "command": "npx",
      "args": ["@backboard/mcp-adapter"],
      "env": {
        "BACKBOARD_API_KEY": "your-api-key",
        "BACKBOARD_ENDPOINT": "https://api.backboard.io"
      }
    }
  }
}
```

### Step 3 — Enable R-CLI coexistence

For projects that also use R-CLI, configure the shared memory bridge:

```bash
# In your project root
backboard init --adapter ox
```

This creates `.backboard/config.json`:

```json
{
  "adapters": {
    "ox": {
      "type": "mcp",
      "memory_path": ".backboard/memory.db"
    },
    "r-cli": {
      "type": "native",
      "memory_path": ".backboard/memory.db"
    }
  },
  "routing": {
    "default_model": "gpt-4o",
    "fallback_chain": ["gpt-4o", "gpt-4o-mini", "claude-3-haiku"]
  }
}
```

### Step 4 — Verify

```bash
ox mcp list
# backboard — registered, connected

backboard status
# Model: gpt-4o | Memory: connected | R-CLI: synced
```

## Available Tools

| Tool | Description |
|---|---|
| `backboard_route(prompt, opts?)` | Route a prompt through smart model fallback |
| `backboard_memory_read(key)` | Read from shared memory store |
| `backboard_memory_write(key, value)` | Write to shared memory store |
| `backboard_memory_context()` | Get current conversation context |
| `backboard_models()` | List available models and their status |
| `backboard_rcli_sync()` | Sync state with R-CLI workspace |
| `backboard_stats()` | Get usage stats and routing decisions |
| `backboard_config()` | Get current Backboard configuration |

## Usage Examples

### Smart Model Routing

```js
const response = await tools.backboard_route({
  prompt: "Explain quantum computing in simple terms",
  max_tokens: 500,
  temperature: 0.7
});

// Backboard auto-selects the best model based on cost/latency
// falls back to cheaper models if primary is unavailable
```

### Shared Memory Bridge

```js
// ox agent writes context
await tools.backboard_memory_write({
  key: "project.context",
  value: {
    goal: "build a landing page",
    tech_stack: "Next.js 15",
    design_system: "shadcn/ui"
  }
});

From SAI141003/ox, MIT.

More from SAI141003/ox