AI Agents Overview: OpenClaw, MCP, and Authentication

Senpi agents are AI agents that you self-host. They consist of two layers:

OpenClaw is the open-source AI agent framework that Senpi agents run on. It handles the agent's brain - reasoning, memory, cron jobs, and tool execution.

The Senpi MCP is the trading interface. Senpi provides 58 MCP tools that plug into OpenClaw to give your agent access to Hyperliquid trading capabilities (market data, strategy management, position management, audit, and more).

Architecture

One endpoint, one token

• There is only one Senpi MCP endpoint: `https://mcp.prod.senpi.ai/mcp`. There is no separate tracker or analytics endpoint - all 58 tools are at this URL.
• Senpi auth tokens have no permission levels. One valid token gives full access to everything.

Agents never need your private keys

If your agent says it needs your wallet keys or private keys to trade, this is wrong and a sign your AI model is confused. Senpi agents authenticate via the Senpi auth token (SENPI_AUTH_TOKEN), which you set in Railway. Never share your private keys with anyone or any bot.

Adding Senpi to an existing OpenClaw agent. Add Senpi as a remote MCP server in your openclaw.json: set the URL to `https://mcp.prod.senpi.ai/mcp` and add your Senpi auth token in the Authorization header. The token must be the raw value - no `Bearer` prefix. All 58 tools will be available immediately.

Before you start, make sure you've signed up and gotten started on Senpi. For setting up a new agent from scratch, see article 18. For troubleshooting, see article 19.