NabilAziz99

NabilAziz99 / agent-runtime

Public

Python port of Claude Code's agent-runtime architecture, on LangChain

118
8
85% credibility
Found May 26, 2026 at 118 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

A Python framework for building AI assistants that can hold conversations, use tools to read and edit files, search code, and spawn specialized sub-agents to handle complex tasks in parallel — inspired by Claude Code's architecture.

How It Works

1
🔍 Discover the project

You find a Python project that lets you build AI assistants with the same smarts as Claude Code — reading files, searching code, and even spawning helper agents.

2
⚙️ Set up your environment

You install the required pieces and connect your AI service so your assistant can think and respond.

3
💬 Start a conversation

You launch the chat program and start typing requests like you would to any helpful assistant.

4
🤔 Watch it work

Your assistant thinks through your request and automatically uses the right tools — reading files, searching for patterns, or running commands.

5
🚀 Watch it delegate

For big tasks, your assistant can bring in specialized helpers — an explorer to search everywhere, a planner to design solutions, or a general helper for anything else.

6
🛑 Cancel anytime

If you change your mind, you can stop everything mid-task and your assistant winds down cleanly without getting stuck.

✅ Get your results

Your assistant completes the task and gives you a clear answer, whether it worked alone or coordinated multiple helpers.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 118 to 118 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is agent-runtime?

agent-runtime is a Python framework that replicates Claude Code's agent spawning architecture. It lets you build AI agents that can spawn subagents, run tools, and handle multi-turn conversations with proper cancellation support. The system is built on LangChain, so it works with Anthropic, OpenAI, or Google models out of the box. You get a Session object for the main agent loop, a registry to track running agents, and a dispatcher that batches tool calls intelligently.

Why is it gaining traction?

The hook is hierarchical cancellation. If you press ESC, the main agent and all its subagents wind down cleanly without leaving orphaned processes. The agent registry acts as a shared mailbox where background subagents push notifications when they finish, keeping the parent loop alive until results arrive. It also mirrors Claude Code's three built-in subagent types (general-purpose, Explore, Plan) so you can spawn specialized agents for file search or architecture planning without reinventing the wheel.

Who should use this?

Backend developers building agentic pipelines who want Claude Code-style spawning without starting from scratch. Researchers prototyping multi-agent workflows will appreciate the clean extension points for hooks and compaction. Anyone running LangChain who needs structured subagent management rather than ad-hoc tool calling.

Verdict

This is early-stage infrastructure (Step 1 of a 6-step roadmap), so expect missing pieces and evolving APIs. The 0.85% credibility score reflects a small, new project with limited community feedback. If you need production-ready agent orchestration today, look elsewhere. If you're comfortable with bleeding-edge tools and want to shape the architecture as it grows, this is worth watching.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.