Zhenfy

Tools

Tools give agents the ability to interact with external systems. Each tool is an MCP (Model Context Protocol) server — a standardized interface for exposing capabilities to AI agents.

How tools work

You provide the MCP server URL when configuring the agent. MCP tools are self-describing — each tool tells the LLM what it does and what parameters it accepts. The agent's system prompt tells it WHEN to use tools; the tool descriptions tell it HOW.

Adding tools to an agent

curl -s -X POST https://zhenfy.ai/api/agent/v1/agents \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Web Researcher",
    "system_prompt": "You are a research assistant. Use your web search tool to find information.",
    "tools": [
      {"url": "https://mcp.example.com/web-search"}
    ]
  }'

Authentication

If your MCP server requires authentication, provide secrets alongside the URL. Secrets are stored securely by the platform and injected at runtime — they never appear in the agent's context or logs.

"tools": [
  {
    "url": "https://mcp.example.com/github",
    "secrets": {"Authorization": "secret:my-github-token"}
  }
]

What makes a good MCP server?

An MCP server exposes one or more tools over HTTP. Each tool has a name, description, and parameter schema. Any API can be wrapped as an MCP server. Common examples:

  • Memory / knowledge base — store and retrieve information across sessions
  • Document store — search and retrieve from your docs, FAQ, or product database
  • Code analysis — clone repos, scan for vulnerabilities, run static analysis
  • Web search — find information on the internet
  • Messaging — send messages to Slack, Discord, or email

See the use cases section for examples of MCP servers paired with system prompts for specific scenarios.

← Agent ConfigurationSub-Agents →