ZenBrain
The neuroscience-inspired memory system for AI agents.
7 memory layers. Real neuroscience — FSRS, Hebbian, sleep consolidation, emotional tagging, plus 10 advanced research modules (vmPFC-FSRS, two-factor Hebbian, simulation-selection sleep, Fiedler-value KG health, IB budget, Hopfield STM, ...).
Pure TypeScript. Zero dependencies. 528 tests. Extracted from a production AI platform.
Status: pre-1.0, semver — the public API can still change before 1.0.
528 tests green on Node 22, 24 and 26 in CI · every release published to npm with build provenance ·
every change recorded in the CHANGELOG · issues and pull requests get a first response typically within 72 hours.
▶ Try the live playground in your browser — runs this published code, no install required.
📄 Paper & Citation — ZenBrain: A Neuroscience-Inspired 7-Layer Memory Architecture for Autonomous AI Systems
- arXiv preprint (cs.AI): arxiv.org/abs/2604.23878
- Open-access archive (Zenodo / CERN): doi.org/10.5281/zenodo.19353663
- ORCID: 0009-0001-1793-012X
- License: CC BY 4.0 (paper) · Apache-2.0 (code)
@misc{bering2026zenbrain,
title = {ZenBrain: A Neuroscience-Inspired 7-Layer Memory Architecture for Autonomous AI Systems},
author = {Bering, Alexander},
year = {2026},
eprint = {2604.23878},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
doi = {10.5281/zenodo.19353663},
url = {https://arxiv.org/abs/2604.23878}
}
Feedback, replications, and counter-results are explicitly welcome — please open an issue or reach out via research@zensation.ai.
Your AI forgets everything after every conversation. ZenBrain fixes that — with the same mechanisms your brain uses: spaced repetition, emotional consolidation, Hebbian strengthening, and exponential forgetting curves. Not a vector database with a wrapper. Actual neuroscience.
Architecture vs. this package. ZenBrain's architecture is 15 neuroscience-inspired mechanisms — 9 foundational algorithms + 6 Predictive Memory Architecture (PMA) components (paper). The 6 PMA components are proprietary and run in the production system. This open-source package ships the algorithm library: 10 core algorithms + 10 advanced research modules (20 modules), zero-dependency.
Benchmark: LongMemEval-500
On LongMemEval-500, ZenBrain wins all nine head-to-head answer-quality comparisons against Letta, Mem0 and A-Mem — three competitors x three LLM judges, under Bonferroni-corrected significance (alpha = 0.05/18, p_min = 6.2e-31, d in [0.18, 0.52]). It reaches 91.3% of a full-context oracle's binary-judge accuracy at 1/106th of the per-query token cost (47.7% vs. 52.2%).
The paper prints where ZenBrain loses as well: on LoCoMo, substring-based aggregate F1 favours lexical retrieval (BM25) by metric design, and we do not contest that. The advantage is most pronounced on judge-graded answer quality and cross-session reasoning.
The mechanism comparison further down re-runs from this repository in under a minute —
bash scripts/compare-mechanisms.sh, no API keys and nothing to install. It prints a positive
and a negative control before the result, so the instrument can be checked before its output
is trusted. The method, the effect sizes and the ablations behind the numbers above are in the
paper; this repository ships no runner for them. The archived packages below run the significance
tests and effect sizes in full, and the mechanism ablation behind paper Tables 7–9.
- Method, effect sizes and ablations: arXiv:2604.23878
- Open-access archive: 10.5281/zenodo.19353663
Reproduction packages
The raw material behind the numbers above is deposited on Zenodo, open access and citable. Both links are concept DOIs and resolve to the latest version, the same convention this README uses for the paper archive; each description was measured against the version named after it.
- Mechanism ablation, paper Tables 7–9 — 10.5281/zenodo.22162063
(described here: v1.0.0). Four experiment suites (95 tests), the reference JSON the paper's
tables were generated from, and
verify-against-reference.mjs, which diffs a fresh run against that reference and exits non-zero on drift.npm install && npm run experiments; the run itself needs no API keys and no network, and finishes in under a minute on a laptop. Two of the paper's other ablation tables need data this package does not carry: Table 11 the LoCoMo corpus, Table 13 a different pipeline. The package says so itself. - Measurement package, LongMemEval-500 and the real-pipeline flag ablation —
10.5281/zenodo.22161977 (described here: v1).
Per-(system, judge, seed) judged outputs, the flag manifests as recorded at run time, a
SHA256SUMS.txtcovering every file in the package, and the analysis scripts. Three of those scripts are stdlib-only and self-checking — the oracle comparison behind the 91.3% figure, the judge-agreement figures, and the real-pipeline flag-ablation table: each prints every re-derived value next to the reference it has to match, and exits non-zero on mismatch. The significance tests behind the nine head-to-head comparisons against Letta, Mem0 and A-Mem sit in a separate script that needs numpy and scipy; it recomputes all eighteen pairwise tests and rewrites the deposited significance JSON byte-identically, so what catches a mismatch there is the checksum, not an exit code.
Both packages name what they do not cover. Replications and counter-results are welcome: research@zensation.ai.
How ZenBrain differs from Mem0, Letta and Zep
ZenBrain implements fifteen mechanisms taken from human memory research. No system among those surveyed in the paper integrates more than two of them. The table below records which of the mechanisms appear in the public source of three widely used memory systems, at pinned versions, on a fixed date.
| Mechanism | ZenBrain | Mem0 | Letta | Zep |
|---|---|---|---|---|
| FSRS spaced repetition | yes | — | — | — |
| Hebbian learning | yes | — | — | — |
| Ebbinghaus forgetting curves | yes | — | — | — |
| Sleep consolidation | yes | — | — | — |
| Emotional tagging | yes | — | — | — |
| Zero runtime dependencies | yes | — | — | — |
How this was measured, 27 August 2026. Full-text search over the
checked-out public source of mem0ai/mem0
(npm mem0ai 3.1.7, PyPI mem0ai 2.0.19),
letta-ai/letta-code
(npm @letta-ai/letta-code 0.31.2) and
getzep/zep, lockfiles excluded. A dash means
the term does not occur in that snapshot — not that the system cannot do
something comparable under another name. Dependency counts are declared direct dependencies:
@zensation/core resolves to two packages, both our own; mem0ai
declares four, @letta-ai/letta-code eighteen. Re-run the whole check yourself with
scripts/compare-mechanisms.sh; it
prints its own positive and negative controls so you can see the instrument works before you
trust the result.
Human memory does not work like a key-value store. The brain keeps specialised systems for different kinds of memory, forgets actively, modulates by emotion and retrieves by context. ZenBrain brings those mechanisms to AI agents.
Advanced algorithms (since v0.3.0, May 2026)
On top of the 10 core algorithms above, @zensation/algorithms ships 10 advanced algorithms grounded in recent neuroscience and ML research. Each is exposed as its own sub-path (@zensation/algorithms/<name>) and remains zero-dependency:
fsrs-vmPFC— Prediction-Error coupled FSRShebbian-two-factor— Two-Factor synaptic consolidationsleep-simulation-selection— RL-based replay selectionspectral-health— Fiedler-value KG health monitorib-budget— Information-Bottleneck retention budgetdopamine-routing·hopfield-stm·personalized-pagerank·surprise-gradient-memory·temporal-multi-route
See CHANGELOG.md for details.
Quick Start
Requires Node.js 22 or newer (since
0.4.0). On Node 20 or older, npm silently installs the last compatible release (@zensation/algorithms@0.3.4,@zensation/core@0.2.2) instead of the current one — which looks like a broken package but is a platform mismatch. See CHANGELOG.
npm install @zensation/algorithms
import {
initFromDecayClass,
getRetrievability,
updateAfterRecall,
tagEmotion,
computeEmotionalWeight,
computeHebbianStrengthening,
propagateForRelation,
} from '@zensation/algorithms';
// 1. Schedule a memory with FSRS
const memory = initFromDecayClass('normal_decay');
// 2. A week later, check recall probability (Ebbinghaus curve)
const aWeekLater = new Date(Date.now() + 7 * 24 * 60 * 60 * 1000);
const retention = getRetrievability(memory, aWeekLater);
console.log(`Recall probability: ${(retention * 100).toFixed(1)}%`);
// ~36.8% — retrievability has decayed over the week
// 3. User recalled it anyway — update scheduling
const updated = updateAfterRecall(memory, 4, retention, aWeekLater);
// stability 7 -> 8.19: recalling at low retrievability gives a bigger boost
// 4. Tag emotional significance
const emotion = tagEmotion('I am absolutely thrilled — I got the promotion!');
const weight = computeEmotionalWeight(emotion);
console.log(`Decay multiplier: ${weight.decayMultiplier}x`);
// 2.7x — emotional memories decay nearly 3x slower
// 5. Strengthen knowledge connections (Hebbian)
const stronger = computeHebbianStrengthening(1.0);
// 1.09 — "neurons that fire together wire together"
// 6. Propagate confidence through your knowledge graph
const confidence = propagateForRelation(0.5, 0.8, 1.0, 'supports');
// 0.9 — supporting evidence increases confidence
Want the advanced algorithms?
import {
computeKGPredictionError,
computeAdaptiveFSRSInterval,
} from '@zensation/algorithms/fsrs-vmPFC';
// Couple FSRS scheduling with the prediction-error signal from your
// knowledge graph: when the embedding has shifted a lot since the last
// review (high cosine distance), shrink the next interval; otherwise push
// it out. Both arrays must have the same length.
const lastEmbedding = [0.1, 0.2, 0.3, 0.4];
const currentEmbedding = [0.5, 0.4, 0.1, 0.2];
const pe = computeKGPredictionError(lastEmbedding, currentEmbedding);
const nextInterval = computeAdaptiveFSRSInterval(14, pe);
Each advanced algorithm has its own sub-path (@zensation/algorithms/spectral-health, @zensation/algorithms/ib-budget, …). All zero dependencies.
Runnable examples
Five self-contained examples live in examples/:
| Example | Shows |
|---|---|
basic-chatbot.ts | Working Memory + Short-Term Memory for conversation context — no SDK needed |
with-claude.ts | An Anthropic Claude assistant that remembers across conversations |
with-langchain.ts | ZenBrain as the memory backend of a LangChain agent |
with-crewai.ts | Multiple agents sharing Working Memory, with Hebbian strengthening |
with-vercel-ai.ts | A memory-aware system prompt for the Vercel AI SDK streamText pattern |
npx tsx examples/basic-chatbot.ts
The integration examples additionally need their respective SDK installed. Want a LlamaIndex.TS or Mastra example? Those are open as good first issues.
The Science Behind It
7-Layer Memory Architecture
Layer 7: Cross-Context Memory ← Shared knowledge across domains
Layer 6: Core Memory ← Pinned facts (Letta-style)
Layer 5: Procedural Memory ← "How to do X" (skills & workflows)
Layer 4: Long-Term Semantic ← Facts with FSRS scheduling
Layer 3: Episodic Memory ← Concrete experiences & events
Layer 2: Short-Term / Session ← Current conversation context
Layer 1: Working Memory ← Active task focus (7±2 items)
Each layer has different retention characteristics, consolidation rules, and retrieval mechanisms — just like the human brain.
FSRS Spaced Repetition
FSRS (Free Spaced Repetition Scheduler) outperforms SM-2 by 30%. It uses the desirable difficulty principle: reviewing when retention is low gives a bigger stability boost. Your AI reviews important facts at optimal intervals — never too early (wasteful), never too late (forgotten).
Emotional Memory
The amygdala modulates memory consolidation — emotional events are remembered more vividly (flashbulb memory). ZenBrain's emotional tagger assigns arousal, valence, and significance scores using a 400+ keyword lexicon (English & German). Emotional memories get up to 3x longer decay half-lives.
Hebbian Learning
"Neurons that fire together wire together" (Hebb, 1949). Knowledge graph edges that are frequently co-activated grow stronger. Unused edges decay and eventually get pruned. The result: a self-organizing knowledge structure that reflects actual usage patterns, with homeostatic normalization to prevent runaway growth.
Ebbinghaus Forgetting Curves
Ebbinghaus (1885) showed that memory decays exponentially: R = e^(-t/S). ZenBrain implements personalized decay profiles that adapt to individual learning patterns, with SM-2 compatibility for existing spaced repetition systems.
Context-Dependent Retrieval
Tulving's Encoding Specificity Principle (1973): memories are recalled better when the retrieval context matches the encoding context. ZenBrain captures temporal context (time of day, day of week) and task type at encoding time, providing up to a 30% retrieval boost when contexts match.
Bayesian Confidence Propagation
Knowledge isn't isolated — facts support or contradict each other. ZenBrain propagates confidence through your knowledge graph using Bayesian belief updates: supporting evidence increases confidence, contradictions decrease it, with damping for numerical stability.
Sleep Consolidation
During sleep, the hippocampus replays recent experiences, strengthening important memories and pruning weak connections (Stickgold & Walker, 2013). ZenBrain simulates this process: selectForReplay() prioritizes emotional and recently-accessed memories, simulateReplay() boosts their stability by 50%, and pruneWeakConnections() removes weak Hebbian edges — implementing the Synaptic Homeostasis Hypothesis (Tononi & Cirelli, 2006).
import { selectForReplay, simulateReplay } from '@zensation/algorithms/sleep-consolidation';
// Select memories for overnight consolidation
const toReplay = selectForReplay(allMemories);
// Simulate sleep replay — stability ↑, weak edges pruned
const result = simulateReplay(toReplay);
console.log(`Replayed ${result.summary.totalReplayed} memories, avg stability +${result.summary.avgStabilityIncrease.toFixed(1)} days`);
Memory Coordinator
The MemoryCoordinator orchestrates all 7 layers into a single cohesive system — inspired by Global Workspace Theory (Baars, 1988):
import { MemoryCoordinator } from '@zensation/core';
const memory = new MemoryCoordinator({ storage: adapter, embedding: embedder });
// Auto-routes to the right layer (semantic, episodic, procedural, or core)
await memory.store('User prefers TypeScript', { type: 'auto' });
// Cross-layer search with ranked, deduplicated results
const results = await memory.recall('programming preferences');
// Consolidate: promote episodic → semantic, apply decay
await memory.consolidate();
// FSRS review queue across all layers
const dueItems = await memory.getReviewQueue();
Packages
| Package | Description | Status |
|---|---|---|
@zensation/algorithms · source | 20 algorithm modules — 10 core (FSRS, Hebbian, Ebbinghaus, emotional, Bayesian, sleep consolidation, intervals, visualization) + 10 advanced (vmPFC-FSRS, two-factor Hebbian, IB budget, Hopfield STM, …) | :white_check_mark: Published |
@zensation/core · source | Memory layers, coordinator, adapter interfaces | :white_check_mark: Published |
@zensation/adapter-postgres · source | PostgreSQL + pgvector storage adapter | :white_check_mark: Published |
@zensation/adapter-sqlite · source | SQLite storage adapter (zero-config) | :white_check_mark: Published |
@zensation/mcp · source | MCP server — gives any MCP client (Claude Desktop, Claude Code, Cursor) the seven layers as four tools. Carries the protocol SDK, so the core stays dependency-free | :white_check_mark: Published |
@zensation/ai-sdk · source | Vercel AI SDK middleware — recall before the model call, store after it. Works with any provider, zero runtime dependencies | :white_check_mark: Published |
Tree-Shakeable Imports
Every algorithm is available as a subpath export:
// Import everything
import { tagEmotion, updateAfterRecall } from '@zensation/algorithms';
// Or just what you need (better tree-shaking)
import { updateAfterRecall } from '@zensation/algorithms/fsrs';
import { tagEmotion } from '@zensation/algorithms/emotional';
import { computeHebbianStrengthening } from '@zensation/algorithms/hebbian';
import { propagateForRelation } from '@zensation/algorithms/bayesian';
import { selectForReplay } from '@zensation/algorithms/sleep-consolidation';
import { getRetrievabilityWithCI } from '@zensation/algorithms/intervals';
import { generateRetentionCurve } from '@zensation/algorithms/visualization';
Use Cases
AI Chatbots with Long-Term Memory
import { updateAfterRecall, getRetrievability, scheduleNextReview } from '@zensation/algorithms/fsrs';
import { tagEmotion, computeEmotionalWeight } from '@zensation/algorithms/emotional';
// When your AI learns a fact about the user:
function rememberFact(fact: string) {
const memory = initFromDecayClass('normal_decay');
const emotion = tagEmotion(fact);
const weight = computeEmotionalWeight(emotion);
// Emotional facts get longer retention
return {
...memory,
emotionalWeight: weight.consolidationWeight,
decayMultiplier: weight.decayMultiplier,
};
}
// Before each conversation, check what needs reinforcement:
function getFactsDueForReview(facts: MemoryState[]) {
return facts.filter(f => getRetrievability(f) < 0.7);
}
Knowledge Graph with Self-Organizing Edges
import { computeHebbianStrengthening, computeHebbianDecay } from '@zensation/algorithms/hebbian';
import { propagateForRelation } from '@zensation/algorithms/bayesian';
// When two concepts are mentioned together:
function coActivate(edge: { weight: number }) {
edge.weight = computeHebbianStrengthening(edge.weight);
}
// Periodic maintenance — decay unused edges:
function decayEdges(edges: { weight: number; lastUsed: Date }[]) {
for (const edge of edges) {
edge.weight = computeHebbianDecay(edge.weight);
// Edges below MIN_WEIGHT (0.1) can be pruned
}
}
RAG with Confidence Scoring
import { propagateForRelation, isSignificantChange } from '@zensation/algorithms/bayesian';
// After retrieval, propagate confidence through related facts:
function updateConfidenceGraph(facts: Fact[], relations: Relation[]) {
for (const rel of relations) {
const newConf = propagateForRelation(
rel.target.confidence,
rel.source.confidence,
rel.weight,
rel.type // 'supports' | 'contradicts' | 'related_to'
);
if (isSignificantChange(newConf, rel.target.confidence)) {
rel.target.confidence = newConf;
}
}
}
Extracted From Production
These aren't toy implementations — ZenBrain's algorithms are extracted from ZenAI, a production AI platform. Everything claimed here is verifiable in this repository:
- 528 tests (429 algorithms + 99 core), all passing
- Zero runtime dependencies — pure TypeScript, dual ESM + CJS, tree-shakeable subpath exports
- Reproducible — building from this source produces the same 153-file
@zensation/algorithms@0.4.4tarball published on npm - 7-layer memory architecture grounded in published neuroscience
Contributing
We welcome contributions! See CONTRIBUTING.md for guidelines. Issues and pull requests get a first response typically within 72 hours.
Resources: API Reference | Recipes | Architecture | Benchmarks | FAQ | Roadmap
# Clone the repo
git clone https://github.com/zensation-ai/zenbrain.git
cd zenbrain
# Install dependencies
npm install
# Run tests
npm test
# Build all packages
npm run build
Research
ZenBrain's architecture and algorithms are documented in an open-access technical disclosure:
- arXiv preprint (cs.AI): arxiv.org/abs/2604.23878
- Open-access archive: ZenBrain: A Neuroscience-Inspired 7-Layer Memory Architecture for Autonomous AI Systems (Zenodo, DOI: 10.5281/zenodo.19353663 — resolves to the latest version)
- TDCommons: Technical Disclosure (CC BY 4.0)
- HuggingFace: Model Card & Benchmarks
If you use ZenBrain in academic work, please cite:
@misc{bering2026zenbrain,
title = {ZenBrain: A Neuroscience-Inspired 7-Layer Memory Architecture for Autonomous AI Systems},
author = {Bering, Alexander},
year = {2026},
eprint = {2604.23878},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
doi = {10.5281/zenodo.19353663},
url = {https://arxiv.org/abs/2604.23878}
}
Community
- GitHub Discussions: Ask a question, show what you built — help, show-and-tell, feature requests
- GitHub Issues: Bug reports & feature requests
- Email: open-source@zensation.ai
License
Apache 2.0 — use it in production, modify it, distribute it. Just keep the attribution.
Built by ZenSation in Kiel, Germany.