Getting started

Build AI agents with persistent intelligence

AI agents need continuity and connected knowledge, but unrestricted history creates privacy, relevance, deletion, and authority risks. Neutron AI gives your application a model-independent memory, knowledge, and decision-intelligence layer: store permissioned memories, represent temporal relationships, retrieve bounded context, compare viable options, observe outcomes, and keep provider credentials and final authority in your own application.

These docs are written for external developers integrating Neutron AI into their own platforms. They cover public APIs, SDKs, CLI workflows, MCP usage, authentication, memory and graph concepts, decision lifecycles, and safe integration patterns.

Browse the documentation

Connect a model provider

Choose a model provider integration for OpenAI, Azure OpenAI, Anthropic Claude, Google Gemini, Vertex AI, Groq, Cohere, Amazon Bedrock, or an OpenAI-compatible endpoint.

Build consequence-aware agents

Use Memory + Consequence Agents for worked workflows, industry use cases, and consequence decision support.

The concrete integration surface

Scoped memory and knowledge

Organise memory, entities, relationships, decisions, and outcomes by Nucleus and Scope so user, project, workspace, team, session, and agent knowledge stays separated.

Context Engine

Retrieve bounded context for a task, including eligible memory, temporal graph state, earlier decisions, consequence evidence, observed outcomes, contradictions, procedures, provenance, and reviewed learning.

Decision guidance

Persist objectives, hard and soft constraints, options, predicted consequences, selected actions, observed outcomes, and decision-time knowledge snapshots. Expose feasibility, Pareto trade-offs, uncertainty, and optional solver proof boundaries without fabricating a universal optimum.

SDKs and CLI

Use public clients for app integration, automation, local validation, CI workflows, and provider connector setup.

MCP integration

Expose a focused set of memory, graph, context, decision, outcome, and simulation tools to compatible MCP clients and agent runtimes through a public MCP endpoint.

Provider guides

Connect Neutron AI to OpenAI, Anthropic, Gemini, Mistral, xAI, DeepSeek, Groq, Cohere, Perplexity, OpenRouter, Azure OpenAI, Amazon Bedrock, and Vertex AI.

Who these docs are for

Product teams adding persistent knowledge and decision support to AI features
Platform teams standardising agent memory, graph, and governance across products
Developers integrating REST, SDK, CLI, or MCP workflows
Agent builders who need reusable user, project, and tool context
Security and governance reviewers validating public integration behavior

Integration path

Start with the Quickstart.
Review Authentication.
Choose the API, SDK, CLI, or MCP path.
Follow the provider guide for your model host.
Learn the Memory model and knowledge and decision model.
Add agent workflows with the Agents and Code Nuclei docs.

Public surfaces

REST API for memory, knowledge, context, decisions, outcomes, learning, and simulations
SDKs for common application environments
CLI for local and operational workflows
Focused MCP tools for compatible agent clients
Provider adapters that keep provider credentials in your application

Safety model

Neutron AI is designed so your application controls authentication, Nucleus and Scope selection, provider calls, solver or simulation adapters, and sensitive data handling. Do not send secrets, credentials, unnecessary personal data, hidden prompts, or raw chain-of-thought as knowledge content.