NEURONAGENT/Agent Runtime

NeuronAgent

REST API and WebSocket agent runtime system with long-term memory, tool execution, and streaming responses. Integrates seamlessly with NeuronDB.

Start Read docs GitHub

curl

Quickstart

curl http://localhost:8080/api/v1/health

REST APIWebSocketLong-term MemoryTool Execution

NeuronAgent

Agent Runtime

Architecture

AI agent runtime architecture with state management, memory store, and tool execution

NeuronAgent Architecture

API Layer

REST API | WebSocket | Health Check

Agent State Machine

• State Persistence
• Task Execution
• Recovery & Retry
• State Transitions

Session Management

• Session Lifecycle
• Context Management
• Message History
• Multi-tenancy

Tool Registry

• SQL Tools
• HTTP Tools
• Code Execution
• Shell Commands

Memory Store

• HNSW Vector Search
• Context Retrieval
• Long-term Memory
• Semantic Search

Job Queue

• Background Jobs
• Retry Logic
• Poison Handling
• SKIP LOCKED

NeuronDB Integration

Vector Search | Embeddings | LLM Operations | PostgreSQL

NeuronAgent Runtime

Build autonomous agents with memory and tool execution

NeuronAgent

console • demo

v1.0ready

Create & manage agents

REST API for agent lifecycle management

HTTP

Agent Creationdemo

POST /api/v1/agents
{
  "name": "support-agent",
  "system_prompt": "You are a helpful support agent...",
  "model": "gpt-4",
  "tools": ["sql", "http", "search"],
  "memory_config": {
    "enabled": true,
    "max_context": 10
  }
}

// Response
{
  "id": "agent_123",
  "status": "active"
}

Results5 rows

id	category
1	agent_123: support-agent, status: active	agent
2	agent_456: sales-agent, status: active	agent
3	agent_789: support-agent, status: idle	agent
4	agent_101: support-agent, status: active	agent
5	Total agents: 4, Active: 3	summary

Performance

Query Time

8.42ms

Latency (P95)

12.5ms

QPS

8.2k

Status

ready

Query Statistics

Execution

Rows Returned5

Cache Hit96%

PlanOptimized

Connection

Databaseneuronagent

Versionv1.0

Statusactive

Summary

Total Queries1,247

Success Rate99.8%

Avg Latency8.2ms

Agent API Docs

Agent Runtime Features

Why NeuronAgent

Agent State Machine

•Complete state machine for autonomous task execution
•State persistence, recovery, and transition management
•Supports complex workflows with conditional branching
•State validation and error handling
•Multi-agent collaboration with shared state management
•Workflow engine for orchestrating multi-step agent tasks

Long-term Memory

•HNSW-based vector search for context retrieval
•Semantic search across historical conversations and knowledge base
•Memory persistence across sessions
•Temporal relevance weighting for context ranking

Tool System

•Extensible tool registry: SQL, HTTP, Code, and Shell operations
•Tool execution with timeout, validation, and error handling
•Streaming responses for long-running operations
•Custom tool registration API

REST API & WebSocket

•Full CRUD API for agents, sessions, and messages
•WebSocket support for real-time streaming responses
•RESTful endpoints with OpenAPI documentation
•Rate limiting and authentication middleware

Background Jobs

•PostgreSQL-based job queue with worker pool
•Retries and poison message handling
•SKIP LOCKED for concurrent processing
•Crash recovery and job state persistence
•Tenant-aware job isolation

Authentication & Security

•API key-based authentication with rate limiting
•Role-based access control (RBAC)
•Request validation and sanitization
•Audit logging for security compliance
•Multi-tenant isolation at the database level

Session Management

•Session lifecycle management with context preservation
•Message history tracking and pagination
•Multi-agent session support
•Session timeout and cleanup policies

NeuronDB Integration

•Deep integration with NeuronDB for embeddings generation
•Vector search and LLM operations
•Direct SQL tool access to NeuronDB functions
•Efficient context retrieval using HNSW indexes

Multi-Agent Collaboration

•Multi-agent collaboration with shared state and communication
•Agent-to-agent messaging and coordination
•Distributed task execution across multiple agents
•Shared memory and context across agent sessions

Workflow Engine

•Workflow engine for orchestrating complex multi-step tasks
•Conditional branching and parallel execution
•Task dependencies and scheduling
•Workflow state management and recovery

Human-in-the-Loop (HITL)

•Human-in-the-loop integration for approval workflows
•Interactive prompts for user feedback and guidance
•Manual intervention points in automated processes
•Human review and approval mechanisms

Budget Management

•Budget management with cost tracking per agent and session
•Token usage monitoring and limits
•API call budget enforcement
•Cost analytics and reporting

Evaluation Framework

•Evaluation framework for agent performance assessment
•Metrics collection: success rate, latency, quality scores
•A/B testing capabilities for agent configurations
•Automated testing and benchmarking tools

Capabilities

Agent runtime features

Capability	Description	Performance	Production Ready
Agent State Machine	State persistence, recovery, transitions	Sub-second state changes	✓
Long-term Memory	HNSW vector search, semantic retrieval	Millisecond context lookup	✓
Tool Registry	SQL, HTTP, Code, Shell tools	Concurrent execution	✓
REST API	Full CRUD operations for agents, sessions	High-throughput requests	✓
WebSocket	Real-time streaming responses	Low-latency streaming	✓
Background Jobs	Job queue with retries, SKIP LOCKED	Non-blocking async ops	✓
Multi-Agent Collaboration	Shared state, agent-to-agent messaging	Distributed task execution	✓
Workflow Engine	Multi-step task orchestration, conditional branching	Parallel execution support	✓
Human-in-the-Loop (HITL)	Approval workflows, interactive prompts	Manual intervention support	✓
Budget Management	Cost tracking, token limits, API call budgets	Real-time cost analytics	✓
Evaluation Framework	Performance metrics, A/B testing, benchmarking	Automated testing tools	✓
Authentication	API keys, rate limiting, RBAC	OAuth-ready	✓

Get Started

Build AI Agents with NeuronAgent

Deploy NeuronAgent. Create autonomous agent systems with persistent memory, tool execution, and streaming responses.

View Documentation View on GitHub