Tools

Agentmd provides three types of tools for your agents.

Built-in Tools

Always available, no configuration needed:

• file_read — Read files from workspace
• file_write — Write files to allowed paths
• file_edit — Edit files with targeted text replacement
• file_move — Move or rename files
• file_glob — Find files matching a pattern
• http_request — Make HTTP calls (GET, POST, etc.)
• memory_save / memory_append / memory_retrieve — Long-term memory
• skill_use / skill_read_file / skill_run_script — Skills (when enabled)
• run_agent — Delegate to other agents (when agents configured)

Learn more →

Custom Tools

Extend with Python. Use the SDK for sandbox-safe file access:

# workspace/agents/_config/tools/my_tool.py
from langchain_core.tools import tool
from agent_md.sdk import resolve_path

@tool
def my_tool(path: str) -> str:
    """Read a file safely."""
    resolved, error = resolve_path(path)
    if error:
        return f"ERROR: {error}"
    return resolved.read_text()

Create custom tools →

MCP Integration

Use external MCP servers:

mcp:
  - fetch
  - github

MCP integration →