rebuno 0.2.0

Python SDK for the Rebuno execution runtime

Rebuno Python SDK

Python client library for the Rebuno agent execution runtime.

Installation

pip install rebuno

Quick Start

from rebuno import RebunoClient

client = RebunoClient(base_url="http://localhost:8080")

# Create an execution
result = client.create_execution(agent_id="my-agent", input={"task": "hello"})
print(result["execution_id"])

Building an Agent

from rebuno.agent import BaseAgent

class MyAgent(BaseAgent):
    def process(self, ctx):
        result = ctx.invoke_tool("web.search", {"query": "hello"})
        return {"answer": result}

agent = MyAgent(
    agent_id="my-agent",
    kernel_url="http://localhost:8080",
)
agent.run()

Building a Runner

from rebuno.runner import BaseRunner

class MyRunner(BaseRunner):
    def execute(self, tool_id, arguments):
        if tool_id == "web.search":
            return {"results": ["..."]}
        raise ValueError(f"Unknown tool: {tool_id}")

runner = MyRunner(
    runner_id="my-runner",
    kernel_url="http://localhost:8080",
    capabilities=["web.search"],
)
runner.run()

MCP Support

Connect to MCP servers to expose their tools through the kernel:

pip install rebuno[mcp]

import asyncio
from rebuno import AsyncBaseRunner

runner = AsyncBaseRunner(
    runner_id="mcp-tools",
    kernel_url="http://localhost:8080",
)
runner.mcp_server(
    "filesystem",
    command="npx",
    args=["-y", "@modelcontextprotocol/server-filesystem", "/tmp"],
)
asyncio.run(runner.run())

Agents can also use MCP tools as local tools:

agent.mcp_server("filesystem", command="npx", args=["-y", "@modelcontextprotocol/server-filesystem", "/tmp"])

See the full documentation for details on partial failure tolerance, config-based setup, and runner MCP routing.

Documentation

See the full documentation.

License

MIT