Skip to content

What is Hadron

Hadron is a platform that provides memory to AI applications. It gives AI agents, chatbots, and automations a persistent, structured memory that survives across sessions, can be shared between applications, and is accessible via MCP tools and APIs.

What Hadron does

Memory for AI

Every AI application needs to remember things: who the user is, what was discussed, what decisions were made, what data was collected. Most AI applications lose this information when the session ends, or store it in ad-hoc formats that can't be shared or searched.

Hadron provides a structured memory graph where AI applications store and retrieve knowledge. The memory is:

  • Persistent — survives across sessions, days, months
  • Structured — nodes with typed content, metadata, and relationships (edges)
  • Searchable — by keywords, tags, node types, or semantic meaning
  • Shareable — across applications, agents, and organizations
  • Encryptable — user data can be encrypted at rest with user-controlled keys

Tools for memory

Hadron connects memory to external tools so AI agents can act on what they know:

  • Email integration — read and respond to email, store conversations in memory
  • Document upload — ingest documents into the memory graph (planned)
  • Web research — discover and store information from the web (planned)
  • MCP tools — a growing set of tools that any MCP-compatible AI agent (Claude Code, Cursor, etc.) can use to read, write, and search memory

Sharing and collaboration

Memory in Hadron is organized by organizations. Within an organization:

  • Multiple memories hold different kinds of knowledge (domain knowledge, user profiles, agent configurations)
  • Memories can be shared across organizations via subscriptions (read-only or read-write)
  • A marketplace (planned) lets organizations publish and discover public memories

No-code agents

Hadron lets you build AI agents without writing code:

  • Define an agent with memories, a system prompt, and surfaces (MCP, API, or both)
  • Attach knowledge memories with domain content the agent can reference
  • Bundle one or more agents inside an app — the unit that exposes agents to the outside world (workstation client like Claude Code, chatbot embed, automation, IoT device, etc.) and issues the API keys callers use to reach them

Chatbot design

A special class of agents — chatbots — use Hadron's conversation engine to drive structured, multi-stage conversations with users:

  • Conversations define the flow (onboarding, diagnosis, action planning)
  • Stages within conversations have prompts, extraction specs, and routing edges
  • Goal-based routing matches the user's intent to the right conversation at every turn
  • Per-user memory stores each user's chat history, extracted data, and goals separately

This is covered in depth in the Chatbot Design section.

Who uses Hadron

  • Developers building AI-powered applications that need persistent memory
  • Chatbot designers creating structured conversational experiences
  • Organizations that want to share domain knowledge across AI agents
  • Researchers using AI agents for deep-dive exploration of complex topics

How to get started