Antikythera Agent Runtime

Multi-agent orchestration with MCP tool integration. Powered by WebAssembly.

Installation

pip install antikythera-agent

Quick Start

Setup with PromptManager

from antikythera_agent import Agent, PromptManager

# Initialize with built-in prompts
prompts = PromptManager()

# Use a built-in prompt
agent = Agent(
    provider="openai",
    model="gpt-4o",
    system_prompt=prompts.get_content("coder")
)

result = agent.run("Write a sorting algorithm")
print(result.output)

Custom Prompts

from antikythera_agent import Agent, PromptManager, PromptConfig

prompts = PromptManager()

# Register custom prompts
prompts.register(PromptConfig(
    id="my-reviewer",
    name="Security Reviewer",
    content="You are a security-focused code reviewer.",
    tags=["reviewer", "security"]
))

# Use custom prompt
agent = Agent(
    provider="openai",
    model="gpt-4o",
    system_prompt=prompts.get_content("my-reviewer")
)

Multi-Agent with Centralized Prompts

from antikythera_agent import Orchestrator, PromptManager, AgentProfileConfig

prompts = PromptManager()

orchestrator = Orchestrator(execution_mode="auto", max_concurrent_tasks=4)

orchestrator.register_agent(AgentProfileConfig(
    id="coder",
    name="Code Writer",
    role="developer",
    system_prompt=prompts.get_content("coder")
))

orchestrator.register_agent(AgentProfileConfig(
    id="reviewer",
    name="Code Reviewer",
    role="reviewer",
    system_prompt=prompts.get_content("reviewer")
))

result = orchestrator.dispatch("Write and review code")

Built-in Prompts

Requirements

• Python >= 3.9

License

MIT