Agenspy
Protocol-First AI Agents for Next Generation Agentic Systems
Build smarter AI agents with protocol intelligence on top of DSPy. Support for MCP, A2A, and future agentic protocols with automatic connection management.
Quick Start
Get started with Agenspy in seconds
pip install agenspy
pip install "agenspy[mcp]"
import dspy
from agenspy import create_mcp_pr_review_agent
# Configure DSPy
lm = dspy.LM('openai/gpt-4o-mini')
dspy.configure(lm=lm)
# Create protocol-aware agent
agent = create_mcp_pr_review_agent("mcp://github-server:8080")
# Use the agent
result = agent(
pr_url="https://github.com/org/repo/pull/123",
review_focus="security"
)
print(f"Review: {result.review_comment}")
print(f"Status: {result.approval_status}")
Why Protocol-First?
Agenspy brings protocol intelligence to DSPy agents, enabling seamless integration with modern agentic systems
Protocol-First Architecture
Built-in support for MCP, Agent2Agent, and future protocols with automatic connection management
DSPy Integration & Optimization
Seamless integration with DSPy framework for optimized agent performance and prompt engineering
Full-featured CLI & Servers
Complete command-line interface with Python/Node servers for development and production
Automatic Connection Management
Handle complex protocol connections, session state, and error recovery automatically
Future Protocol Extensibility
Designed to support WebSocket, gRPC, and emerging protocols as they develop
Session State Management
Persistent session handling across protocol connections with automatic state recovery
Protocol Intelligence Architecture
How Agenspy bridges DSPy agents with protocol ecosystems
Protocol-First Architecture
How Agenspy enables seamless protocol integration
DSPy Agent Layer
Protocol Intelligence
MCP Support
Model Context Protocol for tool integration
A2A Protocol
Agent-to-Agent communication
Future Protocols
WebSocket, gRPC, and more
MCP Integration Example
See how Agenspy seamlessly integrates with Model Context Protocol
from agenspy import MCPAgent
import dspy
# Initialize with automatic connection management
agent = MCPAgent(
server_uri="mcp://github-server:8080",
capabilities=["tools", "resources"]
)
# DSPy integration with protocol intelligence
class PRReviewAgent(dspy.Module):
def __init__(self):
super().__init__()
self.review = dspy.ChainOfThought("context, pr_url -> review, status")
def forward(self, pr_url, context="security"):
return self.review(context=context, pr_url=pr_url)
# Protocol-first usage
review_agent = PRReviewAgent()
result = review_agent(pr_url="https://github.com/org/repo/pull/123")
CLI Commands
Powerful command-line interface for agent development and testing
agenspy agent run
Run an agent with real MCP server integration
agenspy agent run "Review PR https://github.com/stanfordnlp/dspy/pull/8277" --real-mcp
agenspy protocol test
Test protocol server connectivity and capabilities
agenspy protocol test mcp
agenspy demo github-pr
Run the GitHub PR review demo
agenspy demo github-pr --pr-url https://github.com/org/repo/pull/123
Example Projects
Ready-to-run examples to get you started
basic_mcp_demo.py
ViewSimple MCP agent implementation with DSPy
comprehensive_mcp_demo.py
ViewFull-featured agent with advanced capabilities
github_pr_review.py
ViewGitHub pull request review agent
multi_protocol_demo.py
ViewExperimental multi-protocol support
python_server_demo.py
ViewPython MCP server example
Resources & Links
Build the Future of Agent Communication
Join the protocol-first revolution. Start building agents that communicate seamlessly across any protocol.