bigchuidw3

bigchuidw3 / spring-ai-agentx

Public

基于原生 Spring AI 的智能体（Agent）开发框架，提供 ReAct 执行引擎、分层记忆、工具调度、Human-in-the-Loop 等核心能力，形成完整的 Harness Engineering 方案，帮助开发者快速构建 AI Agent，可以快速搭建 Java 版的 Claude Code。

20
1
89% credibility
GitGems finds repos before they trend -- Star growth, AI reviews, and architecture deep-dives -- free with GitHub.
Sign Up Free
AI Analysis
Java
AI Summary

Spring AI AgentX is a Java framework that helps developers build intelligent AI assistants. It provides a complete engine for agents that can reason step-by-step, use tools like file editing and code execution, remember conversations across sessions, and pause to ask humans for input when needed. The framework works with any AI model that supports tool calling, handles both instant and streaming responses, and manages memory in three layers — short-term for the current chat, user profiles for personalization, and long-term semantic memory for knowledge retrieval. It is built on Spring AI and designed to be simple enough that anyone familiar with Spring can use it without learning new concepts.

How It Works

1
💡 You discover a smarter way to build AI assistants

You find Spring AI AgentX while looking for a Java framework to build AI agents that can think, use tools, and remember conversations.

2
🔧 You connect your AI model in three lines of code

You pick your favorite AI model (like DeepSeek, Qwen, or any other supported model) and wire it up with just a few lines of simple setup.

3
🤖 Your AI assistant comes to life and starts working

The framework automatically handles the thinking-and-acting loop, calling tools when needed, pausing to ask you questions, and keeping track of everything.

4
Your assistant can work in two different ways
⚡
Instant answers

Get the full response immediately when you just need the result

📺
Live streaming

Watch the thinking and output appear piece by piece for transparency

5
🧠 Your assistant remembers things across conversations

Short-term memory handles the current chat, while long-term memory stores what it learns about you so it gets smarter over time.

6
⏸️ Your assistant can pause and ask you questions

When it encounters something important or risky, it stops and waits for your input before continuing — keeping you in control.

🎉 Your AI project is ready to ship

You have a fully capable AI agent with memory, tools, and human oversight — built entirely in Java with Spring.

Sign up to see the full architecture

5 more

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 spring-ai-agentx?

Spring AI AgentX is a Java framework for building AI agents, positioned as a "Claude Code for Java." It provides a ReAct (Reasoning + Acting) execution engine that handles multi-turn conversations, tool calling, and memory management out of the box. Built on Spring AI and Reactor, it supports both streaming and synchronous responses while integrating with models like DeepSeek, Qwen, and GLM. The framework manages conversation history, long-term memory, and user profiles automatically through database-backed storage.

Why is it gaining traction?

The main appeal is that it stays close to Spring AI's native APIs rather than inventing new abstractions. Developers familiar with Spring AI can start building agents in three lines of code. The framework handles the messy parts: tool execution, pause/resume flows for human approval, context compression to control token costs, and compatibility with thinking models that output reasoning separately from answers. The built-in DeepSeek V4 adapter solves a real compatibility issue that Spring AI's official modules don't handle well.

Who should use this?

Java backend developers who want to add agent capabilities to Spring applications without learning LangGraph or other Python-centric patterns. Teams already invested in Spring Boot who need structured outputs, database-backed memory, or human-in-the-loop approval workflows will find the most value. Early-stage projects willing to accept some risk could move quickly with it.

Verdict

This is a feature-complete v1.0.0-M1 with solid documentation and a clear design philosophy. The credibility score sits at 0.9%, reflecting the early stage and small community. With only 20 stars, limited test coverage, and no Maven Central release yet, production use requires caution. Worth watching for the M2 release cycle.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.