Neuron AI — Agentic Framework for PHP
Build production-ready AI-driven applications in PHP with agents, workflows, and multi-agent orchestration.
- Source repo
- neuron-core/neuron-ai
- Stars
- ★ 2.1k
- Last updated
- 3d ago
- License
- MIT
- Primary language
- PHP
- FA score
- 45/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformOpenAI API · Claude API
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A PHP developer wants to add an AI assistant to an existing Laravel or Symfony app without rewriting infrastructure.
- Not a fit if
- Projects still on PHP 8.0 or older
- Teams wanting RAG out of the box without configuring a vector store and embeddings provider
- Source review
- 45/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
Neuron is a PHP framework for creating and orchestrating AI Agents, providing a powerful and flexible architecture for integrating AI entities into PHP applications. It covers the entire agentic development lifecycle: LLM interfaces, data loading, multi-agent orchestration, monitoring, and debugging. The framework includes an extendable Agent class, RAG abstractions, and a Workflow engine that allows developers to build everything from simple chatbots to complex automated systems. It supports multiple LLM providers, MCP connectors for tool integration, and emphasizes event-driven workflows with checkpointing and human-in-the-loop patterns. Neuron is designed for teams needing deep AI integration in existing PHP codebases.
Neuron lets you install via Composer and generate an Agent class using the make:agent command. You extend the Agent base class, implementing provider() to specify the LLM provider (e.g., Anthropic, OpenAI), instructions() to set the system prompt, and tools() to register tools or toolkits like MySQLToolkit. The Agent automatically manages memory, function calls, and RAG. Workflow components allow you to compose LLM, embeddings, data loaders, and vector stores as standalone modules into custom pipelines, with support for human-in-the-loop interruptions. Structured output is achieved via PHP attributes (SchemaProperty). Monitoring and debugging are integrated with Inspector.
- A PHP developer wants to add an AI assistant to an existing Laravel or Symfony app without rewriting infrastructure.
- A data team needs to generate reports from a SQL database; the agent uses MySQLToolkit to query and summarize data.
- An ops team wants to automate a workflow with approval steps; Human-in-the-loop supports manual checks.
- A developer needs to extend agent capabilities by connecting external MCP servers (e.g., modelcontextprotocol/server-everything).
- A company wants to build a RAG application using Pinecone vector store and Voyage embeddings to retrieve private docs.
- A developer needs to switch between multiple LLM providers without changing agent logic.
How do you install or deploy this agent?
Requires PHP ^8.1. Install via Composer: composer require neuron-core/neuron-ai.
How do you use this agent?
Run vendor/bin/neuron make:agent DataAnalystAgent to generate an agent class, then edit it to set provider() and instructions(). Instantiate and call: $agent = DataAnalystAgent::make(); $response = $agent->chat(new UserMessage("Hi"))->getMessage(); For monitoring, set the INSPECTOR_INGESTION_KEY environment variable.
What are this agent's strengths and limitations?
- Purpose-built for PHP, deep integration with the PHP ecosystem, offers a Laravel-like developer experience.
- Supports a wide range of LLM providers (Anthropic, OpenAI, Ollama, etc.) with easy switching.
- Built-in MCP connector to integrate external tools, enhancing extensibility.
- Workflow component supports human-in-the-loop, ideal for complex processes needing human approval.
- Provides structured output and RAG support, covering various application needs.
- Requires PHP 8.1+, which may exclude older PHP projects.
- Monitoring relies on the third-party Inspector service, adding an external dependency.
- Framework is relatively young; community and documentation may be less mature than established PHP frameworks.
- Advanced features (MCP, RAG) require additional configuration and components, increasing the learning curve.
How does this agent compare with similar options?
The README mentions it is the only vertical ecosystem for agentic applications in PHP, but does not name specific alternatives.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Neuron AI — Agentic Framework for PHP This agent | 45 · Major gaps | Library / SDKFree + model costs | ★ 2.1k | 3d ago | PHP | OpenAI API · Claude API |
| AgentScope 2.0 | 70 · Some gaps | Library / SDKFree + model costs | ★ 33k | today | Python | OpenAI API · Claude API |
| Synapse AI — Multi-Agent Orchestration Platform | 70 · Some gaps | CLIFree + model costs | ★ 329 | 2mo ago | Python | OpenAI API · Claude API |
| Open Swarm | 40 · Major gaps | — | ★ 829 | 15d ago | JavaScript | Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence: Framework provides tools and MCP connector, but no explicit least-privilege design; human-in-the-loop pattern exists but not mandatory user confirmation; data flow transparency limited, monitoring mentioned but not detailed; sensitive data handling not explicit; dependency security not mentioned; external effects (e.g., DB operations) not restricted; no rollback mechanism; source attribution only via copyright. Deductions: lack of implementation details and documentation support.
Evidence: README and code examples consistent, but actual execution not verified; many external services (DB, vector stores) but no availability guarantees; error messages not detailed. Deductions: static review cannot verify execution, error handling documentation insufficient.
Evidence: Targets PHP developers, provides multiple scenarios (RAG, Workflow, etc.); capability boundaries not explicit; trigger precision not detailed; environment requirement only PHP 8.1, but external services needed. Deductions: capability boundaries and trigger conditions documentation insufficient.
Evidence: README structure clear, install instructions simple; naming stable but no version history; examples abundant but FAQ missing; known limitations not mentioned; MIT license clear; version changelog not provided; maintenance responsibility not explicit. Deductions: missing FAQ, known limitations, and changelog.
Evidence: Output usability high, structured output provided; marginal value high, unique features; cost-benefit not assessed. Deductions: cost-benefit lacks data.
Evidence: Claims in README have documentation links, but no independent verification; cross-source corroboration insufficient; facts and inferences not clearly separated. Deductions: lack of independent verification and fact/inference separation.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Static review cannot verify actual execution; all scores based on source code and documentation.
- No malicious code found, but be aware of external service dependencies and permission management.
- Consult official documentation for more details, especially security and error handling.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
Fix MySQL toolkit foreign key introspection - PR #688 by @cironapoView on GitHub ↗