MCP (external tools)

MCP (external tools)

zames can use tools from MCP servers. The flagship example is @playwright/mcp: it gives the agent a real browser (navigate, click, snapshot, type, …) on top of the one zames already uses for the DeepSeek, Qwen or GLM chat.

Drop a config file (same shape as Claude Code / Cursor):

• ~/.zames/mcp.json - global
• <project>/.zames/mcp.json - project-scoped (later files win)
• <project>/.mcp.json - the common MCP name

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["-y", "@playwright/mcp@latest", "--headless", "--isolated"]
    }
  }
}

IMPORTANT: keep the MCP browser isolated. @playwright/mcp defaults to the SAME profile directory as zames (~/.zames/profile). If it is launched without –isolated (or without its own –user-data-dir), the MCP browser and the agent browser fight over one profile and the DeepSeek chat shows “Something went wrong when opening your profile”. Always pass –isolated as in the example above. Servers can also be remote (“url”: “https://…”, “transport”: “sse”). Their tools show up in the agent as server__tool (e.g. playwright__browser_navigate) and are listed with /mcp and in /status. A server that fails to connect is skipped with a warning and never breaks the agent.

Recommended servers (verified)

A few small, high-value servers tested end to end with zames (on the DeepSeek and Qwen backends; the MCP layer is provider-agnostic, so they work with GLM too):

• Context7 (@upstash/context7-mcp) — fetches CURRENT, version-specific documentation for a library before the agent writes code, instead of relying on a stale training snapshot. Tools: resolve_library_id then query_docs. Verified live on DeepSeek AND Qwen: the agent called both tools and got the current Zod API. This is the one to add first.

  {
    "context7": {
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp@latest"]
    }
  }
  

• Sequential Thinking (@modelcontextprotocol/server-sequential-thinking) — one sequentialthinking tool that gives the model an explicit, revisable plan before it acts; useful for long multi-step tasks.

• Memory (@modelcontextprotocol/server-memory) — a local knowledge graph (9 tools) for durable facts across chats. NOTE: zames already has MEMORY.md

  • AGENTS.md; add this only if you want a queryable graph.

Keep the stack SMALL: every server loads its tool definitions into the model’s context. Add one when you feel its absence, not before.