ainative-mcp-builder
# AINative MCP Builder Guide
## What is an MCP Server?
Model Context Protocol (MCP) servers expose tools that AI agents (Claude Code, Cursor, Windsurf, etc.) can call directly. AINative's MCP servers (zerodb-mcp-server, zerodb-memory-mcp) are built this way.
## Python — FastMCP
```bash
pip install fastmcp
```
```python
# my_mcp_server.py
from fastmcp import FastMCP
import requests
mcp = FastMCP("my-tools")
API_KEY = "ak_your_key"
BASE = "https://api.ainative.studio"
@mcp.tool()
def get_user_credits() -> dict:
"""Get the current user's credit balance."""
return requests.get(
f"{BASE}/api/v1/public/credits/balance",
headers={"X-API-Key": API_KEY}
).json()
@mcp.tool()
def search_memory(query: str, limit: int = 5) -> dict:
"""Search agent memory semantically."""
return requests.post(
f"{BASE}/api/v1/public/memory/v2/recall",
headers={"X-API-Key": API_KEY},
json={"query": query, "limit": limit}
).json()
@mcp.tool()
def store_memory(content: str, memory_type: str = "episodic") -> dict:
"""Store a fact or event in agent memory."""
return requests.post(
f"{BASE}/api/v1/public/memory/v2/remember",
headers={"X-API-Key": API_KEY},
json={"content": content, "memory_type": memory_type}
).json()
if __name__ == "__main__":
mcp.run()
```
```bash
python my_mcp_server.py
```
## Node.js — MCP SDK
```bash
npm install @modelcontextprotocol/sdk
```
```typescript
// server.ts
import { Server } from '@modelcontextprotocol/sdk/server/index.js';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
const server = new Server(
{ name: 'my-mcp-server', version: '1.0.0' },
{ capabilities: { tools: {} } }
);
server.setRequestHandler('tools/list', async () => ({
tools: [{
name: 'get_credits',
description: 'Get current credit balance',
inputSchema: { type: 'object', properties: {} }
}]
}));
server.setRequestHandler('tools/call', async (request) => {
if (request.params.name === 'get_credits') {
const resp = await fetch('https://api.ainative.studio/api/v1/public/credits/balance', {
headers: { 'X-API-Key': process.env.AINATIVE_API_KEY! }
});
return { content: [{ type: 'text', text: JSON.stringify(await resp.json()) }] };
}
});
const transport = new StdioServerTransport();
await server.connect(transport);
```
## Configure in Claude Code
```json
// .claude/mcp.json
{
"mcpServers": {
"my-tools": {
"command": "python",
"args": ["my_mcp_server.py"],
"env": { "AINATIVE_API_KEY": "ak_your_key" }
}
}
}
```
For a published npm package:
```json
{
"mcpServers": {
"my-tools": {
"command": "npx",
"args": ["my-mcp-package"],
"env": { "AINATIVE_API_KEY": "ak_your_key" }
}
}
}
```
## SKILL.md Format for ClawHub
Every MCP tool should have a matching skill file so agents know when to call it:
```markdown
---
name: my-tool-name
description: One-line description. Use when (1) scenario, (2) scenario, (3) scenario.
---
# Tool Name
Brief description and usage examples.
```
Place in `.claude/skills/my-tool-name/SKILL.md`.
## Publish to npm
```bash
# package.json
{
"name": "my-mcp-server",
"version": "1.0.0",
"bin": { "my-mcp-server": "./dist/server.js" },
"main": "./dist/server.js"
}
npm publish
```
## References
- `zerodb-mcp-server/` — Full 76-tool example (Node.js)
- `zerodb-memory-mcp/` — Lightweight 6-tool example (Node.js)
- `src/backend/app/api/v1/endpoints/zerodb_mcp.py` — Backend tool handlers
- MCP spec: `https://modelcontextprotocol.io`
标签
skill
ai
