NeuroCore

Pluggable, YAML-driven framework for building agentic AI applications.

NeuroCore is the chassis for agentic AI. It wires together workflow orchestration, discoverable skills, structured configuration, and a developer-friendly CLI — so you can focus on building intelligent agents, not plumbing.

Install

pip install neurocore-ai

Quick start

neurocore init my-agent
cd my-agent
neurocore run blueprints/agent.flow.yaml

Contents

Guides

• NeuroCore Developer Tutorial
  • What You Will Build
  • Table of Contents
  • Part 1: Getting Started
  • Part 2: Building Your First Skill
  • Part 3: Configuration Deep Dive
  • Part 4: Blueprints and Flow Orchestration
  • Part 5: Adding NeuroWeave (Knowledge Graph Memory)
  • Part 6: Chat App — Personal Experiences Collector
  • Part 7: Multi-Provider LLM Chat Agent
  • Part 8: Multi-Agent Parallel Architecture
  • Part 9: Packaging and Distribution
  • Part 10: Deployment and Operations
  • Part 11: Local Models & Provider Injection
  • Part 12: Run History, Replay & Resume
  • Part 13: Human-in-the-Loop Approval Gates
  • Part 14: Templates & MCP Tools
  • Quick Reference
• NeuroCore Agent Quickstart
  • 1. Install (pull from PyPI)
  • 2. The loop (run every time)
  • 3. Author a blueprint (a worker)
  • 4. Exchange data with the worker
  • 5. Give it an LLM (in neurocore.yaml)
  • In-process instead of CLI
• NeuroCore Agent Manual
  • 1. Mental model
  • 2. The supervisor ⇄ worker contract
  • 3. The build loop
  • 4. Designing workflows
  • 5. Exchanging information — details & idioms
  • 6. Human-in-the-loop (approval gates)
  • 7. Providers — give a worker an LLM
  • 8. Skill catalog — your building blocks
  • 9. Authoring a custom skill (when nothing fits)
  • 10. Deploying a worker
  • 11. Operating workers — command & API reference
  • 12. Rules & best practices (checklist)
  • 13. Worked examples
• Concepts
  • Blueprint
  • Skill
  • Context
  • Provider
  • Run
  • How they fit together
• NeuroCore Architecture
  • Table of Contents
  • 1. System Overview
  • 2. Layer Architecture
  • 3. Package Structure
  • 4. Class Diagram
  • 5. Data Models
  • 6. Error Hierarchy
  • 7. Configuration Flow
  • 8. Config Merging
  • 9. Skill Discovery Flow
  • 10. Skill Lifecycle
  • 11. Blueprint Execution Flow
  • 12. CLI Command Flow
  • 13. Plugin Architecture
  • 14. Component Interaction Matrix
  • 15. Key Design Decisions
  • 16. LLM Provider Subsystem
  • 17. Persistence, Runs & Resume
  • 18. Human-in-the-Loop
• Providers
  • Configuration
  • Supported providers
  • Local models, zero cloud keys
  • Using the provider in a skill
• Persistence & runs
  • Configuration
  • CLI
  • Statuses
  • Replay vs resume
  • Programmatic API
• Human-in-the-loop
  • The approval: step
  • Lifecycle
  • How it works
• Authoring skills
  • A local skill
  • Async, LLM, and retries
  • Publishing a skill package
  • Testing tip
• NeuroCore vs LangGraph
  • Category difference
  • Use them together
  • The one-liner
• Changelog
  • [0.4.0] — 2026-06-15
  • [0.3.0] — 2026-06-14
  • [0.2.1] — 2026-04-03
  • [0.2.0] — 2026-04-03
  • [0.1.0] — 2026-02-28
• API Reference
  • chat
  • agent
  • answer
  • neurocore
  • summarize