Content
# Channel3 MCP Server
## Quick Start (Free Tier)
No API key required. Connect directly and start searching:
```
https://mcp.trychannel3.com/
```
The free tier is rate-limited to **10 requests per minute** per IP address. If you hit the limit, you'll receive an error response.
## Unlimited Access
For unlimited usage and affiliate tracking, create an account and add your API key:
1. **Create an account** at [trychannel3.com](https://trychannel3.com) and create an API key to authenticate requests and affiliate purchases to your account.
2. Add the Channel3 MCP to your agent with `?apiKey=<your-api-key>` in the URL as such:
```
https://mcp.trychannel3.com/?apiKey=placeholder
```
### Cursor
Add to `~/.cursor/mcp.json`:
```json
{
"mcpServers": {
"Channel3": {
"url": "https://mcp.trychannel3.com/?apiKey=<your-api-key>"
}
}
}
```
To use the free tier (no API key, rate-limited), omit the `apiKey` parameter:
```json
{
"mcpServers": {
"Channel3": {
"url": "https://mcp.trychannel3.com/"
}
}
}
```
### VS Code
Add to `.vscode/mcp.json`:
```json
{
"servers": {
"Channel3": {
"type": "http",
"url": "https://mcp.trychannel3.com/?apiKey=<your-api-key>"
}
}
}
```
### Claude Code
```bash
claude mcp add --transport http Channel3 "https://mcp.trychannel3.com/?apiKey=<your-api-key>"
```
### Claude Desktop
Add to your Claude Desktop config file:
**macOS:** `~/Library/Application Support/Claude/claude_desktop_config.json`
**Windows:** `%APPDATA%\Claude\claude_desktop_config.json`
```json
{
"mcpServers": {
"Channel3": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.trychannel3.com/sse?apiKey=<your-api-key>"
]
}
}
}
```
### OpenAI Agents SDK (Python)
```python
import asyncio
from agents import Agent, Runner
from agents.mcp import MCPServerStreamableHttp
async def main():
async with MCPServerStreamableHttp(
name="Channel3",
params={
"url": "https://mcp.trychannel3.com/?apiKey=<your-api-key>",
},
cache_tools_list=True,
) as server:
agent = Agent(
name="Shopping Agent",
instructions="You are a personal shopping assistant.",
mcp_servers=[server],
)
result = await Runner.run(agent, "I'm looking for a new laptop")
print(result.final_output)
asyncio.run(main())
```
### OpenAI Agents SDK (TypeScript)
```typescript
import { Agent, run, MCPServerStreamableHttp } from "@openai/agents";
const server = new MCPServerStreamableHttp({
url: "https://mcp.trychannel3.com/?apiKey=<your-api-key>",
name: "Channel3",
});
await server.connect();
const agent = new Agent({
name: "Shopping Agent",
instructions: "You are a personal shopping assistant.",
mcpServers: [server],
});
const result = await run(agent, "I'm looking for a new laptop");
console.log(result.finalOutput);
await server.close();
```
# Local Testing
1. Start the dev server: `pnpm run dev`
2. Start MCP Inspector: `npx @modelcontextprotocol/inspector`
3. In the MCP Inspector UI:
- **Transport Type**: Streamable HTTP
- **URL**: `http://localhost:8787/?apiKey=<YOUR_API_KEY>`
- Click **Connect**
