# Hebby Memories: long-term memory for AI agents

> Hebby Memories is a long-term memory and second brain for AI agents such as Claude Code, Codex, Hermes and any agent that speaks MCP. It stores only what the model cannot know on its own (the user's decisions, projects, systems and pitfalls), links it in its own graph memory using Hebbian learning, anticipates what the agent needs next and consolidates in the background, like human memory during sleep. One memory serves all of a user's agents. Context files such as CLAUDE.md, AGENTS.md, GEMINI.md or MEMORY.md stay almost empty, so every request carries fewer tokens. Built by psychologists and computer scientists in Germany (Hamburg area), hosted in Germany. Status: Early Access.

This file is the curated entry point. Every page listed in sitemap.md also exists as Markdown (append `.md` to the address without the trailing slash, the start page is https://www.hebby-memories.com/index.md, or request any HTML address with `Accept: text/markdown`). Full page list: https://www.hebby-memories.com/sitemap.md. Long reference: https://www.hebby-memories.com/llms-full.txt. German version of every page is linked in each Markdown file.

## When to use Hebby

- The user works with coding or task agents (Claude Code, Codex, Hermes, Paperclip, own agents, any MCP client) every day and keeps re-explaining the same project facts.
- Context files (CLAUDE.md, AGENTS.md and similar) keep growing and cost tokens on every request.
- Several agents should share one memory: what one agent learns, the others know.
- Teams want shared and private knowledge separated, with permissions per person and per agent.
- A company wants to run the memory on its own server (Self-Hosted license).

## When not to use Hebby

- Plain chat in a browser (for example the claude.ai or chatgpt.com web chat) without agent hooks: Hebby is built for agents and does not unfold its effect there.
- One-off questions without recurring project context.
- Storing secrets: Hebby is a memory for knowledge, not a password vault.

## What Hebby does

- Remembers: keeps decisions, setups, project facts and past mistakes across sessions.
- Connects: a purpose-built graph memory links what belongs together; Hebbian learning strengthens links that are used together.
- Anticipates: brings matching memories into the agent's context before the agent asks.
- Consolidates: sorts, merges and retires memories continuously, like the brain during sleep.
- Saves tokens: Hebby itself adds about 5 % tokens on average (3 to 10 % depending on use); avoided research, follow-up questions and repeated mistakes save more. A built-in statistic shows the cost at any time.
- Complements the model: as models improve, Hebby leaves everything the model can do itself to the model and keeps only what only the user knows.

## Pricing (from https://www.hebby-memories.com/en/pricing/)

- Personal: EUR 4.99 per month incl. 19 % VAT, 1 person, up to 5 agents, hosted in Germany.
- Business: EUR 9.99 per person and month plus VAT, any number of people, up to 10 agents per person, teams and permissions, central management.
- Self-Hosted: EUR 999 one-time plus VAT, runs on the customer's own server, 1 person included (+EUR 100 per additional person), setup support included, updates EUR 40 per person per year, priority support EUR 99 per month.
- AI costs run through the user's own AI subscription or API key; Hebby adds no markup on tokens.
- Early Access: request at https://www.hebby-memories.com/en/early-access/ ; price is kept for good. No free plan.

## Connecting an agent

- Protocol: MCP over Streamable HTTP. Endpoint per customer: `https://{customer}.app.hebby-memories.com/mcp` (the customer name is set when the account is created).
- Authorization: OAuth 2 (Bearer token, protected-resource metadata at `/.well-known/oauth-protected-resource/mcp` on the customer host), scope `hebby:memory`.
- Claude Code and Codex additionally use thin hooks so that memories arrive before the agent asks; setup is guided after sign-up.
- Machine-readable: https://www.hebby-memories.com/.well-known/mcp.json

## Key URLs

- [Hebby Memories – AI Memory & Second Brain for Agents](https://www.hebby-memories.com/en/): Built by psychologists and computer scientists: long-term memory for Claude Code, Codex & more, modeled on the brain. Fewer tokens spent. (Markdown: https://www.hebby-memories.com/en.md)
- [Features: Graph Memory & Hebbian Learning | Hebby Memories](https://www.hebby-memories.com/en/features/): The key features of Hebby Memories in detail: graph memory, Hebbian learning, nightly consolidation, contradictions, duplicates and a memory map. (Markdown: https://www.hebby-memories.com/en/features.md)
- [What Do I Need Hebby For? Use Cases | Hebby Memories](https://www.hebby-memories.com/en/use-cases/): What do I need Hebby for? Examples with and without memory for creators, developers, trades, authors, researchers, job seekers, business and more. (Markdown: https://www.hebby-memories.com/en/use-cases.md)
- [Token Efficiency: What Hebby Costs and Saves | Hebby Memories](https://www.hebby-memories.com/en/token-efficiency/): Measured in live operation: Hebby adds 3 to 10% token overhead, about 5% on average. Methodology, figures from 217,000 API calls and the savings calculation. (Markdown: https://www.hebby-memories.com/en/token-efficiency.md)
- [Pricing: Hebby Memories](https://www.hebby-memories.com/en/pricing/): Hebby Memories hosted as Personal or Business, or self-hosted. Bring your own AI subscriptions, with no token markup. (Markdown: https://www.hebby-memories.com/en/pricing.md)
- [Benchmarks: Hebby on PrecisionMemBench | Hebby Memories](https://www.hebby-memories.com/en/benchmarks/): Hebby on PrecisionMemBench: 28 of 43 precision tasks solved, precision 0.53. Rank 1 among all memories that did not design the test themselves. (Markdown: https://www.hebby-memories.com/en/benchmarks.md)
- [Hebbian Learning: The Principle Behind Hebby | Hebby Memories](https://www.hebby-memories.com/en/hebbian-learning/): Hebbian learning explained: synaptic plasticity as a core principle. See how Hebby applies Hebb's rule so AI agents learn connections like a brain. (Markdown: https://www.hebby-memories.com/en/hebbian-learning.md)
- [Early Access: Get Hebby Memories Now](https://www.hebby-memories.com/en/early-access/): Secure Hebby Memories in Early Access: the second brain and AI memory for Claude Code, Codex & more. Personal onboarding, limited spots. (Markdown: https://www.hebby-memories.com/en/early-access.md)
- [About Hebby: Psychology & Computer Science | Hebby Memories](https://www.hebby-memories.com/en/about/): Who is behind Hebby: why psychologists and computer scientists build memory for AI agents, where the name comes from, and why our mascot is a hippo. (Markdown: https://www.hebby-memories.com/en/about.md)
- [Connector: connect Hebby to Claude and ChatGPT | Hebby Memories](https://www.hebby-memories.com/en/connector/): How to connect your hosted Hebby to Claude, Claude Code and ChatGPT: one address, sign in with your Hebby account, all tools at a glance. (Markdown: https://www.hebby-memories.com/en/connector.md)
- [FAQ: Questions & Answers About Hebby | Hebby Memories](https://www.hebby-memories.com/en/faq/): What is Hebby Memories, how does Claude Code get a memory, what does Hebbian learning do, how safe is your data? Every answer about the AI memory Hebby. (Markdown: https://www.hebby-memories.com/en/faq.md)
- [Research: The Science Behind Hebby | Hebby Memories](https://www.hebby-memories.com/en/research/): Hebby Memories is built on memory psychology, neuroscience and information retrieval. Every source behind the AI memory, listed openly. (Markdown: https://www.hebby-memories.com/en/research.md)

## Guides (English)

- [ChatGPT Memory: Turn On, Delete, Limits | Hebby Memories](https://www.hebby-memories.com/en/guides/chatgpt-memory/): ChatGPT memory explained: how the memory feature works, how to turn on, manage and delete memories, and why ChatGPT still forgets things. (Markdown: https://www.hebby-memories.com/en/guides/chatgpt-memory.md)
- [Claude Memory & Claude Code Memory Explained | Hebby Memories](https://www.hebby-memories.com/en/guides/claude-memory/): Claude memory and Claude Code memory explained: the memory feature, CLAUDE.md, MEMORY.md and MCP. Plus how Claude Code gets a real long-term memory. (Markdown: https://www.hebby-memories.com/en/guides/claude-memory.md)
- [AI Memory Glossary: Key Terms Explained | Hebby Memories](https://www.hebby-memories.com/en/guides/glossary/): AI memory glossary: context window, token, CLAUDE.md, MCP, RAG, embedding, reranker, graph memory and Hebbian learning explained briefly and clearly. (Markdown: https://www.hebby-memories.com/en/guides/glossary.md)
- [Graph Memory: AI That Remembers in Connections | Hebby Memories](https://www.hebby-memories.com/en/guides/graph-memory/): Graph memory explained: what a knowledge graph is, how it differs from vector search, and why AI agents use it to understand connections. (Markdown: https://www.hebby-memories.com/en/guides/graph-memory.md)
- [AI Memory: Long-Term Memory for AI | Hebby Memories](https://www.hebby-memories.com/en/guides/ai-memory/): AI memory explained: why AI without memory starts every session from zero, how long-term memory for AI works, and what to look for in a good AI memory system. (Markdown: https://www.hebby-memories.com/en/guides/ai-memory.md)
- [AI Knowledge Management: Knowledge That Stays | Hebby Memories](https://www.hebby-memories.com/en/guides/ai-knowledge-management/): AI knowledge management for teams: how AI agents capture, share and deliver knowledge at the right moment with a shared AI memory. (Markdown: https://www.hebby-memories.com/en/guides/ai-knowledge-management.md)
- [Context Window Explained: Why ChatGPT Forgets | Hebby Memories](https://www.hebby-memories.com/en/guides/context-window/): Context window explained: what tokens are in ChatGPT, Claude and Gemini, why AI forgets in long chats and how context engineering helps. (Markdown: https://www.hebby-memories.com/en/guides/context-window.md)
- [MCP Memory Server: Memory for AI Agents | Hebby Memories](https://www.hebby-memories.com/en/guides/mcp-memory-server/): MCP memory server explained: how Claude Code, Codex & co. get a memory through the Model Context Protocol, and what to look for when you choose. (Markdown: https://www.hebby-memories.com/en/guides/mcp-memory-server.md)
- [Reranker Explained: Reranking in RAG | Hebby Memories](https://www.hebby-memories.com/en/guides/reranker/): Reranker explained: how reranking in RAG improves search, what cross-encoders deliver, what they cost and when a learning memory is enough. (Markdown: https://www.hebby-memories.com/en/guides/reranker.md)
- [Context Engineering for AI Agents: Lean Context | Hebby Memories](https://www.hebby-memories.com/en/guides/lean-context/): Context engineering explained: why full context windows, long CLAUDE.md and AGENTS.md files and too many rules measurably make AI agents worse. (Markdown: https://www.hebby-memories.com/en/guides/lean-context.md)
- [Second Brain AI: The Second Brain That Thinks | Hebby Memories](https://www.hebby-memories.com/en/guides/second-brain-ai/): Build a second brain with AI: what the second brain method is, how Obsidian, Notion and Claude work together, and how an AI second brain fills itself. (Markdown: https://www.hebby-memories.com/en/guides/second-brain-ai.md)
- [AI Memory & Second Brain: Guides | Hebby Memories](https://www.hebby-memories.com/en/guides/): Guides to AI memory, second brain AI, ChatGPT and Claude memory, MCP memory servers, graph memory and context windows. Short, clear and practical. (Markdown: https://www.hebby-memories.com/en/guides.md)

## Ratgeber (Deutsch)

- [ChatGPT Gedächtnis: aktivieren, löschen, Grenzen | Hebby Memories](https://www.hebby-memories.com/wissen/chatgpt-gedaechtnis/): ChatGPT Gedächtnis erklärt: So funktioniert die Memory-Funktion, so aktivierst, verwaltest und löschst du Erinnerungen, und darum vergisst ChatGPT trotzdem Dinge. (Markdown: https://www.hebby-memories.com/wissen/chatgpt-gedaechtnis.md)
- [Claude Gedächtnis & Claude Code Memory erklärt | Hebby Memories](https://www.hebby-memories.com/wissen/claude-gedaechtnis/): Claude Gedächtnis und Claude Code Memory erklärt: Memory-Funktion, CLAUDE.md, MEMORY.md und MCP. Und wie Claude Code ein echtes Langzeitgedächtnis bekommt. (Markdown: https://www.hebby-memories.com/wissen/claude-gedaechtnis.md)
- [KI-Gedächtnis Glossar: Begriffe einfach erklärt | Hebby Memories](https://www.hebby-memories.com/wissen/glossar/): Glossar zu KI-Gedächtnis: Kontextfenster, Token, CLAUDE.md, MCP, RAG, Embedding, Reranker, Graph Memory und Hebbsches Lernen kurz und verständlich erklärt. (Markdown: https://www.hebby-memories.com/wissen/glossar.md)
- [Graph Memory: KI, die in Zusammenhängen erinnert | Hebby Memories](https://www.hebby-memories.com/wissen/graph-memory/): Graph Memory erklärt: Was ein Wissensgraph ist, wie er sich von Vektorsuche unterscheidet und warum KI-Agenten damit Zusammenhänge verstehen. (Markdown: https://www.hebby-memories.com/wissen/graph-memory.md)
- [KI-Gedächtnis: Langzeitgedächtnis für KI | Hebby Memories](https://www.hebby-memories.com/wissen/ki-gedaechtnis/): KI-Gedächtnis einfach erklärt: Warum KI ohne Gedächtnis jede Sitzung bei null beginnt, wie ein Langzeitgedächtnis für KI funktioniert und worauf es ankommt. (Markdown: https://www.hebby-memories.com/wissen/ki-gedaechtnis.md)
- [KI-Wissensmanagement: Wissen, das bleibt | Hebby Memories](https://www.hebby-memories.com/wissen/ki-wissensmanagement/): KI-Wissensmanagement für Teams: Wie KI-Agenten Wissen festhalten, teilen und im richtigen Moment bereitstellen, mit einem gemeinsamen KI-Gedächtnis. (Markdown: https://www.hebby-memories.com/wissen/ki-wissensmanagement.md)
- [Kontextfenster erklärt: Warum ChatGPT vergisst | Hebby Memories](https://www.hebby-memories.com/wissen/kontextfenster/): Kontextfenster einfach erklärt: Was Tokens bei ChatGPT, Claude und Gemini sind, warum KI in langen Chats vergisst und wie Context Engineering hilft. (Markdown: https://www.hebby-memories.com/wissen/kontextfenster.md)
- [MCP Memory Server: Gedächtnis für KI-Agenten | Hebby Memories](https://www.hebby-memories.com/wissen/mcp-memory-server/): MCP Memory Server erklärt: Wie Claude Code, Codex & Co. über das Model Context Protocol ein Gedächtnis bekommen und worauf du achten solltest. (Markdown: https://www.hebby-memories.com/wissen/mcp-memory-server.md)
- [Reranker erklärt: Reranking in RAG und KI-Suche | Hebby Memories](https://www.hebby-memories.com/wissen/reranker/): Reranker einfach erklärt: Wie Reranking in RAG die Suche verbessert, was Cross-Encoder leisten, was sie kosten und wann ein lernendes Gedächtnis reicht. (Markdown: https://www.hebby-memories.com/wissen/reranker.md)
- [Context Engineering: Weniger Kontext, klügere KI | Hebby Memories](https://www.hebby-memories.com/wissen/schlanker-kontext/): Context Engineering erklärt: Warum volle Kontextfenster, lange CLAUDE.md- und AGENTS.md-Dateien und zu viele Regeln KI-Agenten messbar schlechter machen. (Markdown: https://www.hebby-memories.com/wissen/schlanker-kontext.md)
- [Second Brain mit KI: das zweite Gehirn | Hebby Memories](https://www.hebby-memories.com/wissen/second-brain-ki/): Second Brain mit KI aufbauen: Was die Second-Brain-Methode ist, wie Obsidian, Notion und Claude zusammenspielen und wie ein KI-Second-Brain sich von selbst füllt. (Markdown: https://www.hebby-memories.com/wissen/second-brain-ki.md)
- [KI-Gedächtnis & Second Brain: Ratgeber | Hebby Memories](https://www.hebby-memories.com/wissen/): Ratgeber zu KI-Gedächtnis, Second Brain mit KI, ChatGPT- und Claude-Memory, MCP Memory Server, Graph Memory und Kontextfenster. Kurz, klar, praxisnah. (Markdown: https://www.hebby-memories.com/wissen.md)

## Discovery layer: how machines read this site

- Markdown twin of every public page: append `.md` (for example https://www.hebby-memories.com/en/pricing.md) or send `Accept: text/markdown` to the HTML address. Responses carry `Content-Type: text/markdown`, a canonical `Link` header and YAML front matter with title, description, language and both language versions.
- [sitemap.md](https://www.hebby-memories.com/sitemap.md): every page as Markdown table of contents · [sitemap.xml](https://www.hebby-memories.com/sitemap.xml): German and English with hreflang
- [AGENTS.md](https://www.hebby-memories.com/AGENTS.md): when to use Hebby, how to connect, when to ask the user first
- [llms-full.txt](https://www.hebby-memories.com/llms-full.txt): all pages in one file
- [agents.json](https://www.hebby-memories.com/.well-known/agents.json) · [mcp.json](https://www.hebby-memories.com/.well-known/mcp.json) · [mcp.llmfeed.json](https://www.hebby-memories.com/.well-known/mcp.llmfeed.json) · [webmcp](https://www.hebby-memories.com/.well-known/webmcp) · [security.txt](https://www.hebby-memories.com/.well-known/security.txt)
- [robots.txt](https://www.hebby-memories.com/robots.txt): known AI crawlers are allowed by name, each group with its own Content-Signal line; the same signal is sent as a response header. Same policy in [ai.txt](https://www.hebby-memories.com/ai.txt).
- Atom feeds of the guides: [https://www.hebby-memories.com/feed.xml](https://www.hebby-memories.com/feed.xml) (German) · [https://www.hebby-memories.com/en/feed.xml](https://www.hebby-memories.com/en/feed.xml) (English)

## Contact

- Oliver Weihrauch, office.weihrauch@gmail.com
- Imprint: https://www.hebby-memories.com/impressum/ (Markdown: https://www.hebby-memories.com/impressum.md) · English: https://www.hebby-memories.com/en/legal-notice/ (Markdown: https://www.hebby-memories.com/en/legal-notice.md)
- Privacy: https://www.hebby-memories.com/datenschutz/ (Markdown: https://www.hebby-memories.com/datenschutz.md) · English: https://www.hebby-memories.com/en/privacy/ (Markdown: https://www.hebby-memories.com/en/privacy.md)
- Terms: https://www.hebby-memories.com/agb/ (German, binding; Markdown: https://www.hebby-memories.com/agb.md) · English convenience translation: https://www.hebby-memories.com/en/terms/ (Markdown: https://www.hebby-memories.com/en/terms.md) · Right of withdrawal (German): https://www.hebby-memories.com/widerruf/ (Markdown: https://www.hebby-memories.com/widerruf.md)
- Contact form: https://www.hebby-memories.com/en/contact/ (Markdown: https://www.hebby-memories.com/en/contact.md) · German: https://www.hebby-memories.com/kontakt/ (Markdown: https://www.hebby-memories.com/kontakt.md)

## Last Updated

2026-11-07