universal-agent-tools 1.1.0

Reusable tools and utilities for Universal Agent stack

Universal Agent Tools

Reusable tools and utilities extracted from the Universal Agent Nexus examples.

Overview

This package contains production-ready, reusable components for:

• Ollama + Tools Integration - Working pattern for Ollama LLMs with MCP tools
• Code Chunking - Multi-format intelligent chunking (Python, YAML, Markdown, JSON)
• MCP Server Framework - Base classes for creating MCP-compliant servers
• Tool Registry - Centralized tool discovery and management
• Codebase Analyzer - AST parsing, dependency graphs, PageRank scoring
• GitHub Integration - CLI-based GitHub operations
• Document Generator - Structured document generation
• Sync State Management - Generic sync state tracking

Installation

pip install universal-agent-tools

Or install with optional dependencies:

# With Qdrant support
pip install universal-agent-tools[qdrant]

# With GitHub API support
pip install universal-agent-tools[github]

# Development dependencies
pip install universal-agent-tools[dev]

Quick Start

Ollama + Tools

from universal_agent_tools.ollama_tools import (
    MCPToolLoader,
    create_llm_with_tools
)

# Load tools from MCP server
tools = MCPToolLoader.load_from_server("http://localhost:8000/mcp")

# Create LLM with tools
llm, tools = create_llm_with_tools(tools, model="qwen2.5-coder:14b")

Code Chunking

from universal_agent_tools.code_chunking import chunk_content

# Chunk any file type
chunks = chunk_content(content, "example.py")

# Each chunk has: type, name, content, line_start, line_end, docstring
for chunk in chunks:
    print(f"{chunk['type']}: {chunk['name']} (lines {chunk['line_start']}-{chunk['line_end']})")

Tool Registry

from universal_agent_tools.tool_registry import get_registry

registry = get_registry()
registry.register_server("filesystem", "http://localhost:8000/mcp")
tools = registry.discover_tools()

for tool in registry.list_tools():
    print(f"{tool.name}: {tool.description}")

Architecture

This package follows SOLID design principles:

• Single Responsibility - Each module has one clear purpose
• Open/Closed - Extensible without modification (e.g., custom chunkers)
• Liskov Substitution - Protocols/interfaces ensure substitutability
• Interface Segregation - Focused, minimal interfaces
• Dependency Inversion - Depend on abstractions, not concretions

Components

Priority 1 (Production-Ready)

1. ollama_tools/ - Ollama + LangChain + MCP integration
2. code_chunking/ - Multi-format chunking system
3. mcp_framework/ - MCP server framework

Priority 2 (Well-Tested)

4. tool_registry/ - Tool discovery system
5. codebase_analyzer/ - AST analysis and dependency graphs
6. github_tools/ - GitHub CLI integration

Priority 3 (Useful Patterns)

7. document_generator/ - Structured document generation
8. sync_state/ - Generic sync state tracking

Testing

# Run all tests
pytest

# With coverage
pytest --cov=universal_agent_tools

# Specific component
pytest tests/test_chunking.py

License

MIT License

Contributing

This package is extracted from universal_agent_nexus_examples. See REFACTORING_ANALYSIS.md in the examples repo for extraction details.