ACP Integration Examples¶

Examples of using CodeOptiX with Agent Client Protocol (ACP).

Setting Up ACP Agents¶

Registering Agents¶

from codeoptix.acp import ACPAgentRegistry

registry = ACPAgentRegistry()

# Register Claude Code agent
registry.register(
    name="claude-code",
    command="python claude_agent.py",
    description="Claude Code via ACP"
)

# Register custom agent
registry.register(
    name="my-agent",
    command="node my-agent.js",
    cwd="/path/to/agent",
    description="Custom agent implementation"
)

# List all agents
agents = registry.list_agents()
print(agents)

Using Quality Bridge¶

from codeoptix.acp import ACPQualityBridge

# Create quality bridge
bridge = ACPQualityBridge(
    agent_command="python claude_agent.py",
    behaviors="insecure-code,vacuous-tests",
    auto_eval=True
)

# The bridge will automatically evaluate code quality
# as the agent generates code

Multi-Agent Judge Workflow¶

Setting Up Multi-Agent Evaluation¶

from codeoptix.acp import ACPAgentRegistry
from codeoptix.acp.judge import ACPJudge

registry = ACPAgentRegistry()

# Register multiple agents
registry.register("generator", "python claude_agent.py")
registry.register("judge1", "python grok_agent.py")
registry.register("judge2", "python gpt_agent.py")

# Create multi-agent judge
judge = ACPJudge(
    generate_agent="generator",
    judge_agents=["judge1", "judge2"],
    behaviors=["insecure-code", "vacuous-tests"]
)

# Run evaluation
results = await judge.evaluate_prompt(
    prompt="Write a secure user authentication system",
    context={"requirements": "Must use bcrypt, validate inputs, prevent SQL injection"}
)

print(f"Overall Score: {results['overall_score']}")

Editor Integration¶

Zed Editor Integration¶

Start CodeOptiX as ACP agent:
```
codeoptix acp register
```
Configure Zed to connect to CodeOptiX at the displayed endpoint
CodeOptiX will automatically evaluate code quality in real-time

VS Code Integration¶

Use the ACP extension to connect to CodeOptiX agents.

Custom ACP Agent Implementation¶

Basic Agent Structure¶

from acp import Agent, Server
from codeoptix.acp import CodeOptiXAgent

class MyACPAgent(Agent):
    def __init__(self):
        super().__init__()
        self.codeoptix = CodeOptiXAgent()

    async def handle_generate_code(self, prompt: str) -> dict:
        """Handle code generation requests."""
        result = await self.codeoptix.generate_code(prompt)

        return {
            "code": result.code,
            "tests": result.tests,
            "quality_score": result.metadata.get("quality_score", 0.0)
        }

# Start the agent
async def main():
    agent = MyACPAgent()
    server = Server(agent)
    await server.run()

if __name__ == "__main__":
    import asyncio
    asyncio.run(main())

ACP Quality Bridge Example¶

Automated Quality Monitoring¶

from codeoptix.acp import ACPQualityBridge

# Bridge between editor and agent with quality checks
bridge = ACPQualityBridge(
    agent_command="python coding_agent.py",
    behaviors="insecure-code,vacuous-tests,plan-drift",
    auto_eval=True,
    quality_threshold=0.8  # Require 80% quality score
)

# The bridge will:
# 1. Forward requests to the coding agent
# 2. Evaluate generated code against behaviors
# 3. Provide quality feedback to the editor
# 4. Block low-quality code if threshold not met

await bridge.start()