Agent memory infrastructure that never silently loses or overwrites a fact.
CLS++ is the durable, lineage-tracked memory layer for production agents. Every belief change is archived, attributed, and auditable — so you can debug, replay, and trust what your agents remember.
Watch a belief change leave a receipt
This demo runs the real CLS++ archive-with-lineage contract in your browser (no server). Ordinary agent memory overwrites; CLS++ archives, versions, and links.
Agent learns a fact
Starting…
Access Status
Sign up now and get immediate access to CLS++ Memory API — open worldwide.
Memory you can trust
Built so your agent's memory never silently loses or wrongly overwrites a fact — and proves what it did to each one.
Belief revision with lineage
A contradicting fact archives the old one with full lineage and version — never a silent overwrite or delete. Your agent's memory has a provable history.
No silent data loss
Writes persist from the engine's own index; a failed embedder degrades ranking, never loses a fact. Run the proof — it's reproducible.
Drop-in for your agent
Three lines to integrate. Native adapters for CrewAI, LangGraph, LlamaIndex, AutoGen and MCP. TypeScript and Python SDKs.
Durable by design
Writes are fsync-acknowledged to a write-ahead log, with recovery tested against process death and torn-tail records. An accepted write is a kept write.
Model-agnostic recall
Store once, inject into any model provider your agent uses — Claude, GPT, Gemini, or local. Provider switches don't require re-teaching facts.
Yours to control
Encrypted in transit; delete any memory on request (GDPR right-to-be-forgotten). Your data stays yours.
Three lines of code. Memory you can prove.
Install clsplusplus, call learn / ask / correct / history — and get a memory layer with full lineage. No rewrite of your agent stack.
- Archives every superseded fact with lineage — never a silent overwrite
- Survives embedder failures — a dead embedder loses nothing
- Drop-in for CrewAI, LangGraph, LlamaIndex, AutoGen and MCP
- Python and TypeScript SDKs — same package name: clsplusplus
pip install clsplusplus
from clsplusplus import Brain
brain = Brain("my-agent", api_key="cls_live_...")
# Teach a fact
brain.learn("Customer is on the Pro plan")
# Correct it — old belief is archived with lineage
brain.correct(
"Customer is on the Pro plan",
"Customer upgraded to Enterprise",
)
brain.correct(
"Customer is on the Pro plan",
"Customer upgraded to Enterprise",
)
# Current belief (archived history is one query away)
brain.ask("What plan is the customer on?")
brain.history("customer plan") # current + archived with lineageReady for memory your agents can trust?
Start with 500 free API calls per day. Enough to evaluate in production. No credit card required.