Skip to content

Integrate a Framework

ClawBio skills are framework-agnostic by design. This guide explains how to connect your agent framework to the ClawBio skill registry so your agents can discover and execute skills.

Architecture

Your Agent Framework
        │
        ▼
  ClawBio Registry  ──▶  SKILL.md (contract)
        │
        ▼
  Python CLI skill  ──▶  --input / --output / --demo

Each skill exposes a standard CLI interface. Your framework needs to:

  1. Discover skills from the registry
  2. Read the SKILL.md to understand inputs, outputs, and domain rules
  3. Execute the Python script with the correct arguments
  4. Capture the output directory contents

MCP Bridge

ClawBio provides an MCP (Model Context Protocol) server that exposes skills as tools. If your framework supports MCP, this is the simplest integration path:

# Start the MCP bridge
clawbio mcp serve

# Your MCP-compatible agent can now discover and call skills as tools

Direct Integration

If your framework does not support MCP, integrate directly with the CLI:

import subprocess
import json

def run_skill(skill_name, input_path, output_dir):
    """Execute a ClawBio skill and return the output path."""
    result = subprocess.run(
        ["python", f"skills/{skill_name}/{skill_name}.py",
         "--input", input_path,
         "--output", output_dir],
        capture_output=True, text=True
    )
    if result.returncode != 0:
        raise RuntimeError(result.stderr)
    return output_dir

Registry API

The skill registry provides metadata for discovery:

from clawbio import registry

# List all available skills
skills = registry.list_skills()

# Get skill metadata (parsed from SKILL.md frontmatter)
meta = registry.get_skill("pharmgx-reporter")
print(meta["inputs"])   # Expected input formats
print(meta["outputs"])  # What the skill produces

Testing Your Integration

# Verify your framework can discover skills
clawbio list

# Run a demo skill to test the execution path
python3 skills/pharmgx-reporter/pharmgx_reporter.py --demo --output /tmp/test

# Check output was produced
ls /tmp/test/

Questions?