Kronvex is a persistent memory API for AI agents. It lets you store, recall, and inject context across sessions using 3 REST endpoints.

How does Kronvex work with LangChain?

Kronvex provides a custom BaseMemory class for LangChain that persists context across sessions. Install with pip install kronvex and use KronvexMemory in your ConversationChain.

Is Kronvex GDPR compliant?

Yes. Kronvex stores all data in the EU (Frankfurt, Germany) and is GDPR compliant by architecture. No data is sent outside the EU. A GDPR Art. 17 right-to-erasure endpoint is available, and a standard DPA is provided at kronvex.io/dpa.

What is the difference between RAG and agent memory?

RAG retrieves documents from a knowledge base at query time. Agent memory stores and recalls conversation context and user-specific information across sessions. Kronvex provides agent memory — persistent, semantically searchable context that travels with each agent across all sessions.

How much does Kronvex cost?

Kronvex starts free forever (3 agents, 500 memories, no credit card required). Paid plans start at €19/month. Enterprise pricing is custom.

Does Kronvex offer a Data Processing Agreement (DPA)?

Yes. A standard GDPR Art. 28 DPA is available at kronvex.io/dpa. It covers sub-processors (Supabase Frankfurt, Railway EU), data transfers, security measures, and the right to erasure. Included by default on Scale and Enterprise plans.

🇪🇺 EU-hosted · GDPR-native · No credit card required

Your AI agents forget everything. Kronvex remembers.

Add persistent memory to your LangChain, CrewAI or n8n agent in 5 minutes. 3 API calls. <45ms recall. EU-hosted, GDPR-native.

$ npx @kronvex/setup

Try Live Demo →

Free plan · No credit card · Live demo, no signup

<50ms

semantic recall

🇪🇺

EU · GDPR native

€29/mo

starting price

5min

to first memory

Recall score

0.94

preference · 11ms

Memories stored

247

across 2 agents

kronvex_memory.py

import kronvex

agent = kronvex.Agent(
  agent_id="sales-bot-01",
  api_key="kv-your-key"
)

# 1 · Store a memory
agent.remember(
  "User prefers async comms",
  memory_type="preference"
)

# 2 · Semantic recall
agent.recall("communication style")

# 3 · Inject into prompt
ctx = agent.inject_context(
  "Draft outreach email"
)

200 OK · 11ms

{
  "context": "User prefers async comms",
  "memories": [{
    "content": "User prefers async comms",
    "score":   0.94,
    "type":    "preference"
  }]
}

Time to first memory

5 min

1 key · 3 endpoints

Status

API live

99.9% uptime

inject_context

< 50ms

p99 · EU hosted

Active agent

sales-bot-01

Builder · kv-••••4f2a

Memories stored

247

across 2 agents · EU

Souveraineté

🇪🇺 EU · GDPR

Frankfurt · ISO 27001

Data residency

No US transfer

Schrems II · Art.46

Semantic recall

score 0.94

cosine · 11ms

Ⅱ

Features

Built for production
from day one.

Memory types

Clotho · The Spinner Lachesis · The Allotter Atropos · The Inflexible

Every kind of memory,
one place.

Store facts, preferences, decisions and context. Each type indexed separately for precision recall.

Preference

Prefers concise bullet replies over paragraphs

Fact

Company HQ based in Paris, France

Decision

API integration · production stack

Context

Onboarding complete · tour finished

Scale

Millions of memories. Zero friction.

∞

memories on Scale plan

Performance

<50ms DB query

vector search only · ~3-4s end-to-end incl. embedding

Compliance

GDPR processor layer — built in.

🇪🇺

EU hosting, erasure API, DPA, audit trail. You stay in control of consent and data flows.

Security

No LLM at recall

Your users' data never transits through OpenAI when memories are read. Embeddings are generated at write time only — recall is pure pgvector similarity search. No third-party data exposure on every query.

← Unlike Mem0, which routes every read through an LLM

Multi-agent

Isolated agent spaces.

TTL Memory Decay

Memories that age like humans do.

Configure TTL per memory type. Stale context fades. Critical memories persist.

# Expire in 30 days
agent.remember("Trial active", ttl_days=30)

# Pin permanently
agent.remember("VIP customer", pinned=True)

Sessions

Cross-session continuity.

Every memory is tied to a session. Query across all sessions or filter to a specific conversation.

Confidence Score

Not just similarity.
Real relevance.

Every recalled memory is ranked by a composite score — not raw cosine distance. Recent, frequently-accessed memories rise to the top.

Semantic match

How closely the memory matches what the agent is asking — the primary signal.

Recency

A memory from yesterday outweighs one from six months ago. Decay rate adapts to memory type.

Frequency

Memories recalled often are more reliable. The more an agent confirms a fact, the higher it ranks.

Live ranking

0.91 Prefers concise replies — mentioned 3× this week HIGH

0.54 Uses dark mode — set 4 months ago, never re-confirmed MED

0.18 Trial plan — onboarding context from 6 months ago LOW

Ⅱ

recall · "database" 41ms

3 MEMORIES · SORTED BY CONFIDENCE

0.91 NEW Migrated to PostgreSQL 16 last week

0.43 STALE Was using MongoDB

0.31 STALE Evaluating SQLite for local cache

INJECT threshold=0.5 → 1 memory injected · noise filtered automatically

Smart memory — never stale

New context
always wins.

When your agent stores new information that contradicts older memories, Kronvex's confidence scoring automatically surfaces the most relevant, recent context — and filters out stale data before it pollutes your system prompt.

Confidence formula

confidence = similarity × 0.6
+ recency × 0.2
+ frequency × 0.2

60% semantic match

20% recency (sigmoid 30d)

20% access frequency

threshold 0.5 by default

Ⅳ

TOKEN SAVINGS

Stop sending your agent's entire history on every call.

Inject-context returns only memories relevant to the current query — not the full history.

tokens saved

Without Kronvex

0M tokens

$0.00

With Kronvex

0M tokens

$0.00

You Save

0M tokens

$0.00

Estimate: 10M input tokens/month. Without Kronvex, each call includes ~20k tokens of history. With /inject-context, only the top 5 relevant memories are sent (~3.4k tokens) — 83% fewer input tokens.

Ⅴ

Why Kronvex

Memory API.
Actually priced
for developers.

Every memory API charges you to exist. Mem0 jumps from $19 → $249 the moment you need graph memory. Memstate has no GDPR documentation and undisclosed data hosting. Both bill per seat.

Kronvex is for builders: one flat price, three endpoints, running in 5 minutes — hosted in Frankfurt, EU, no lock-in.

EU-hosted · Frankfurt GDPR Free tier Python & Node SDK MCP Server Flat pricing

⚡

5-MIN SETUP

One key, one agent, first memory stored. No devops, no config files.

💶

FLAT PRICING

No per-memory fees. No usage spikes. One price, all features included.

🔌

MCP SERVER

Connect Claude, Cursor, Cline or Windsurf in one npx command. No code required.

📊

CONFIDENCE SCORE

Every recalled memory comes with a relevance score. Filter noise programmatically.

🎯

INJECT CONTEXT

One call returns a ready-to-paste context block. Drop it straight into your LLM prompt.

🔄

AUTO-CONSOLIDATION

Duplicate memories merge automatically. Your agent stays sharp — not bloated.

Feature	KV Kronvex	Mem0	Memstate
Setup time	5 min	~30 min	~5 min
Predictable flat pricing	✓	~	✗
Knowledge graph built-in	✓	✗	✗
EU data residency	✓	✗	✗
🧠Typed memorysemantic · episodic · procedural	✓	✗	✗
🔄Memory TTL / decayUNIQUE	✓	✗	✗
📊Confidence scoringUNIQUE	✓	✗	✗
Inject context endpoint	✓	✗	✗
🔌MCP server	✓	✗	✓
❤️Memory health scoreUNIQUE	✓	✗	✗
🔄Auto-consolidationUNIQUE	✓	✗	✗
Total memory isolation	✓	~	~
GDPR compliant	✓	✓	~
Starting price	Free · €29/mo	Free / $29+	$15+/mo

✓ fully supported · ~ partial · ✗ not available · See full comparison →

Ⅵ

REST API

Three endpoints.
Full docs.
That's it.

REST · JSON SDKs Python · Node.js p99 <100ms

Three POST endpoints. One X-API-Key header.

POST/agents/{id}/remember

POST/agents/{id}/recall

POST/agents/{id}/inject-context

POST/agents

DEL/agents/{id}/memories/{mem_id}

Full API Reference →

import requests

r = requests.post(
  "https://api.kronvex.io/api/v1/agents/agent_001/remember",
  headers={"X-API-Key": "kv-..."},
  json={
    "content": "Prefers email over phone",
    "type":    "preference"
  }
)

curl -X POST \
  https://api.kronvex.io/api/v1/agents/agent_001/remember \
  -H "X-API-Key: kv-..." \
  -H "Content-Type: application/json" \
  -d '{
  "content": "Prefers email over phone",
  "type":    "preference"
}'

const r = await fetch(
  "https://api.kronvex.io/api/v1/agents/agent_001/remember", {
  method: "POST",
  headers: {
    "X-API-Key":     "kv-...",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    content: "Prefers email over phone",
    type:    "preference"
  })
})

Ⅶ

🇪🇺 EU data hosting — Frankfurt, Germany

GDPR-native · Zero-config compliance

Encryption at rest & in transit

Pricing

Flat monthly. No per-seat.
No metered calls.

All plans include the full API. No hidden fees.

Builder

€29/mo

5 agents · 20k memories

For solo devs and indie hackers shipping their first AI agent.

✓
5 agents
✓
20,000 memories
✓
Session filtering
✓
Full REST API + SDKs

POPULAR

Startup

€99/mo

15 agents · 75k memories

For early-stage startups with real users in production.

✓
15 agents
✓
75,000 memories
✓
Session isolation
✓
Audit trail
✓
GDPR basics

Business

€349/mo

50 agents · 500k memories

For growing SaaS with production workloads and compliance needs.

✓
50 agents
✓
500,000 memories
✓
DPA + SLA 99.9%
✓
Custom TTL / decay
✓
Priority support

See all plans →

Builder

€29/mo

5 agents · 20k memories

✓Session filtering
✓Full API + SDKs

POPULAR

Startup

€99/mo

15 agents · 75k memories

✓Session isolation
✓Audit trail
✓GDPR basics

Business

€349/mo

50 agents · 500k memories

✓DPA + SLA 99.9%
✓Custom TTL / decay
✓Priority support

See all plans →

    All prices include VAT (20%) · Non-EU clients VAT-exempt · Monthly billing, cancel anytime
  

FAQ

Can I switch plans at any time?

Yes. Upgrades take effect immediately and are prorated. Downgrades apply at the next billing cycle. No lock-in.

What counts as one "memory"?

One call to /remember stores one memory — a text chunk up to ~2 000 tokens. Recalling and injecting context never count against your quota.

What happens when I hit my memory limit?

New /remember calls return a 429 error. Existing memories and all recall / inject-context endpoints keep working. Upgrade instantly from the dashboard — no data is lost.

Is my data stored in the EU?

Yes. All data is stored in Supabase EU (Frankfurt, Germany). We sign DPAs on request and are GDPR-compliant. No data leaves the EU.

Do you offer a free trial?

The Demo plan is free forever (3 agents, 500 memories). For paid plans, contact us — we can arrange a trial for teams evaluating Kronvex for production use.

Ⅷ

Multi-agent memory

One memory bank.
Every agent.

Each agent has its own isolated namespace, but they can all share a common memory pool. Orchestrator, sub-agents, and tools — all reading from the same memory store.

Isolated namespaces per agent — no memory bleed between tasks

Orchestrator reads any agent's memory via a single API key

Context injected per-agent — scored by similarity, recency, and frequency

agent_swarm.py

# Each agent has its own memory
researcher = kronvex.Agent("researcher_01")
writer     = kronvex.Agent("writer_01")
reviewer  = kronvex.Agent("reviewer_01")

# Researcher stores findings
researcher.remember("User needs TypeScript, EU-hosted")

# Writer and reviewer get same context
writer.inject_context("Draft the proposal")
reviewer.inject_context("Check requirements")

Your AI agents forget everything. Kronvex remembers.

Watch it work.
Real data. Live API.

Built for production
from day one.

Every kind of memory,
one place.

Millions of memories. Zero friction.

<50ms DB query

GDPR processor layer — built in.

No LLM at recall

Isolated agent spaces.

Memories that age like humans do.

Cross-session continuity.

Not just similarity.
Real relevance.

New context
always wins.

Three endpoints.
The whole API.

Stop sending your agent's entire history on every call.

Memory API.
Actually priced
for developers.

Three endpoints.
Full docs.
That's it.

Flat monthly. No per-seat.
No metered calls.

One memory bank.
Every agent.

Resources & guides

Your AI agents forget everything. Kronvex remembers.

Watch it work.Real data. Live API.

Built for productionfrom day one.

Every kind of memory,one place.

Millions of memories. Zero friction.

<50ms DB query

GDPR processor layer — built in.

No LLM at recall

Isolated agent spaces.

Memories that age like humans do.

Cross-session continuity.

Not just similarity.Real relevance.

New contextalways wins.

Three endpoints.The whole API.

Stop sending your agent's entire history on every call.

Memory API.Actually pricedfor developers.

Three endpoints.Full docs.That's it.

Flat monthly. No per-seat.No metered calls.

One memory bank.Every agent.

Resources & guides

Watch it work.
Real data. Live API.

Built for production
from day one.

Every kind of memory,
one place.

Not just similarity.
Real relevance.

New context
always wins.

Three endpoints.
The whole API.

Memory API.
Actually priced
for developers.

Three endpoints.
Full docs.
That's it.

Flat monthly. No per-seat.
No metered calls.

One memory bank.
Every agent.