SULCUS

Memory That Thinks.

The Virtual Memory Management Unit for AI Agents

Your agent forgets everything the moment its context window fills. Sulcus gives it a real memory— a reactive knowledge graph that surfaces what matters, fades what doesn't, and pages the right context in at the right time. Thermodynamic decay. Entity relationships. Interaction-aware recall.

How It Works

We didn't contort the LLM. We accelerated the system around it.

Memories aren't static rows in a database. They have heat — born hot, cooling with time, reheating on recall, spreading warmth through edges to related knowledge. When an agent searches, the engine doesn't just find matches. It ignites the graph — and reactive triggers fire automatically to reinforce, notify, or act.

  • Record — every memory enters the graph at heat 1.0
  • Decay — type-specific half-lives cool memories naturally (24h for episodes, 365d for facts)
  • Recall — searching boosts heat and stability. Spaced repetition makes memories stickier
  • Diffuse — heat spreads through edges to related memories. Recall one, warm many
  • Fold — cold memories consolidate into dense semantic summaries. Nothing is lost

Autonomous Memory Ecosystem

Four specialized components for persistent, intelligent recall.

01

SIU v2 Pipeline

Four subsystems run on every store: SIVU scores utility, SICU classifies type, SILU extracts entities via GPT-5.4-nano and builds the knowledge graph, and SITU evaluates reactive triggers. Memory that understands itself.

02

AGE Knowledge Graph

Apache AGE temporal knowledge graph built on Postgres. 4,131 vertices, 6,713 edges — growing with every store. Cypher queries, entity extraction, temporal traversal. Self-healing on every store and recall.

03

Thermodynamic Engine

Three decay modes: Time-only, Interaction-only, and Hybrid (default). Type-specific half-lives, recall-boost stability, heat-driven resonance. Relevance-weighted recall: similarity × 0.7 + heat × 0.3.

04

Curator System

A background curation cycle acts as the system's sleep cycle — reclassifying, consolidating, summarizing, and re-vectorizing memories to keep the graph accurate and lean as knowledge evolves.

Full Control

30+ Configurable Knobs. Zero Hardcoded Behavior.

Every parameter of the memory engine is exposed via API. Tune decay profiles, resonance depth, trigger rules, consolidation thresholds, and context budgets per tenant, per type, per node.

Active Index
max_nodes
context_budget
hot_threshold
cold_threshold
Resonance
spread_factor
depth
damping
thermal_gate
Reinforcement
on_recall
on_update
on_edge
stability_gain
Consolidation
cold_trigger
cold_threshold
strategy
Tick Mode
interval_ms
trigger_ops
max_idle_ms
Decay Profiles
half_life
floor
reinforce
stab_gain

No Competitor Has This

Reactive Memory Triggers

Set rules on your memory graph. When events happen — a memory is stored, recalled, boosted, or starts to decay — Sulcus fires actions automatically. Pin important memories. Boost fading context. Webhook your systems. The agent doesn't have to remember to remember.

on_storepin

Auto-pin every preference memory so it never decays.

on_recallboost

Reinforce memories every time they're searched — spaced repetition, automated.

on_decaynotify

Alert the agent when critical knowledge starts cooling. Act before it's lost.

on_boostwebhook

Fire HTTP callbacks when memory heat is increased. Chain external systems into your memory lifecycle.

4 events · 7 actions · filter by type, namespace, label pattern, heat threshold · unlimited triggers per tenant

Developer-First

Five lines. Real memory.

Zero-dependency SDKs for Python and Node.js. Or connect directly via MCP or REST.

Pythonpip install sulcus
from sulcus import Sulcus

s = Sulcus(api_key="sk-...")
s.remember("User prefers dark mode",
           memory_type="preference")
results = s.search("dark mode")
Node.jsnpm install @digitalforgestudios/sulcus
import { Sulcus } from "@digitalforgestudios/sulcus";

const s = new Sulcus({ apiKey: "sk-..." });
await s.remember("User prefers dark mode",
  { memoryType: "preference" });
const results = await s.search("dark mode");
Full Documentation →

Privacy-First Architecture

Your memories. Your machine. Your rules.

Most memory systems require your data to live on someone else's server. Sulcus runs locally by default. Cloud sync is optional, encrypted, and you control what leaves the machine.

01

Local-First

The WASM core and MCP sidecar run entirely on your hardware. No network calls required for reads or writes. Your agent's memory never touches a cloud server unless you explicitly enable sync.

02

Zero Knowledge Sync

When you do enable cloud sync, data is encrypted in transit (TLS 1.3) and at rest (AES-256). Tenant isolation ensures your memory graph is invisible to other users — and to us.

03

Data Sovereignty

Self-host the entire stack — server, database, sync — in your own infrastructure. No phone-home telemetry, no surprises. Enterprise licensing available for on-premise deployments.

04

Selective Sync

Choose which namespaces sync to the cloud and which stay local. Sensitive memories (credentials, personal data, health info) can remain on-device while shared knowledge replicates across your fleet.

GDPR-ready · SOC2 roadmap · No telemetry · Enterprise on-premise available

Performance Validated

<25ms
Context Build Time
32
Server Modules
v2.2.1
Current Release
13k
SIU Training Samples
View Documentation

Give Your Agents a Brain.

Free tier. No credit card. Start building agents with real memory in under five minutes.

Create Free Account

No credit card required · Privacy-first · Your data, your control.

Stay in the Loop

Releases, memory research, and what we're building.