Comparison · Agent Memory API
Zep vs Kronvex
EU-hosted, 9× cheaper
Zep shut down their free community edition and moved to cloud-only at $475/mo. Kronvex is a fully managed memory API from €29/mo, EU-hosted, with GDPR compliance built in. Here's an honest comparison for teams evaluating their options.
EU-hosted · Frankfurt · GDPR-native
from €29/mo vs $475/mo
Demo key — 100 memories free
TL;DR
Side-by-side in 30 seconds
| Feature / Question | Zep Cloud | Kronvex |
|---|---|---|
| Data hosting | ✗ US-only, no EU residency | ✓ EU-only — Supabase Frankfurt |
| GDPR compliance | ✗ Not built-in for EU | ✓ Native — erasure, TTL, export |
| Pricing | From $475/mo (Flex+) | From €29/mo (Builder) |
| Free tier | ✗ Community edition deprecated | ✓ Demo key — 1 agent, 100 memories |
| Confidence scoring | ✓ Zep has scoring | ✓ similarity × 0.6 + recency × 0.2 + freq × 0.2 |
| inject-context endpoint | ✗ Build context yourself | ✓ One call returns formatted system prompt |
| Python & Node.js SDK | ✓ Official SDKs | ✓ pip install kronvex · npm install kronvex |
| Time to first memory | ~15–30 min (SDK + session setup) | Under 5 min (POST /remember) |
| Open source | ✗ Community edition deprecated | SDKs open source, API hosted |
| Recall@1 accuracy (open benchmark) | not published | ✓ 87% — methodology |
Data accurate as of March 2026. Sources: getzep.com, kronvex.io/docs.
Honest comparison
Which one to use?
Consider Zep when…
You need graph-based memory extraction
- You need automatic entity and relationship extraction from conversations
- You want LLM-powered memory summarisation built in
- Your team is US-based and EU data residency is not a requirement
- Budget is not a constraint — $475/mo is acceptable for your use case
Use Kronvex when…
You need EU-hosted agent memory at a sane price
- You were using Zep's community edition and need a managed replacement
- GDPR compliance and EU data residency are non-negotiable requirements
- You want predictable flat-rate pricing without surprise usage bills
- You need inject-context — one call returns a formatted system prompt block
- You want to be live in under 5 minutes with a free demo key
- Confidence scoring (recency + frequency weighted) matters to your use case
Code comparison
The same task, two approaches
ZEP — Before (Cloud, $475/mo, US-hosted)# pip install zep-python from zep_python import ZepClient from zep_python.memory import Memory, Message client = ZepClient( api_key="z-...", base_url="https://api.getzep.com" # US-hosted ) # Create a session (required) client.memory.add_session(session_id="user-123") # Store memory (creates graph, calls LLM internally) client.memory.add( session_id="user-123", memory=Memory( messages=[Message( role="user", content="user prefers dark mode" )] ) ) # Recall (no inject-context endpoint) memories = client.memory.get(session_id="user-123") context = "\n".join(m.content for m in memories.messages) # Now format manually and inject into prompt...
KRONVEX — After (€29/mo, EU-hosted, GDPR-native)# pip install kronvex from kronvex import Kronvex client = Kronvex(api_key="kv-...") # One line setup # Store memory — no LLM, pure pgvector embedding await client.agent("user-123").remember( "user prefers dark mode" ) # Recall with confidence scoring memories = await client.agent("user-123").recall( "display preferences" ) # inject-context: one call → formatted system prompt context = await client.agent("user-123").inject_context( "How should I respond?" ) # → "User prefers dark mode (confidence: 0.94)" # EU-hosted. GDPR-native. 9× cheaper.
Why Kronvex instead of Zep
Three reasons teams are switching
9× cheaper
Zep Cloud starts at $475/mo for serious usage. Kronvex Builder is €29/mo — the same core memory functionality (remember, recall, inject-context) at a fraction of the cost. For indie builders and small teams, the economics are not even close.
€29/mo vs $475/moEU data residency
Zep Cloud has no EU data residency option. For teams serving European users, every agent memory crosses the Atlantic. Kronvex stores all data in Supabase Frankfurt. GDPR compliance — right to erasure, memory TTL, data export — is built into the API, not an enterprise add-on.
Frankfurt · GDPR · right to erasureinject-context
Zep returns raw memory objects — you still format the prompt yourself. Kronvex's inject-context endpoint retrieves top memories, ranks them by confidence score, and returns a ready-to-inject system prompt block in a single API call. One fewer step in your agent loop.
one call → ready promptQuestions
Frequently asked
Zep moved to a cloud-only model in 2024, retiring the open-source community edition they previously offered. Their cloud product now requires a subscription starting at $475/mo with no EU data residency. Teams that were using the self-hosted version need a managed alternative. Kronvex is designed exactly for that migration — with a generous free demo tier to get started immediately.
For the core use case — storing agent memories and recalling them semantically — yes. Kronvex covers
remember, recall, and inject-context. The API surface is intentionally simpler (3 endpoints vs Zep's session/graph model). Migration from Zep typically takes less than a day. Kronvex does not replicate Zep's LLM-powered graph extraction, but that feature covers less than 10% of typical agent memory use cases.
Zep Cloud's starting plan (Flex+) is $475/mo. Kronvex Builder is €29/mo — roughly 9–10× cheaper for equivalent core functionality. Both include semantic memory storage, recall, and multi-agent support. Kronvex additionally includes EU hosting, GDPR-native APIs, and inject-context at that price. Zep Cloud has no EU data residency option on any standard plan.
Yes — always. All data is stored in Supabase Frankfurt (eu-central-1, Germany). The API runs on Railway EU. No data ever leaves the EU. GDPR compliance features — right to erasure via
DELETE /memories, per-agent memory TTL, and full data export — are first-class API features, not add-ons. Zep Cloud has no EU option on standard plans.
No. Kronvex stores memories as text and computes embeddings using
text-embedding-3-small (1536 dims). There is no LLM in the write path and no LLM at recall time. Recall uses pure pgvector cosine similarity with our confidence scoring formula: confidence = similarity × 0.6 + recency × 0.2 + frequency × 0.2. This makes recall fast (<40ms p99), predictable in cost, and independent of any OpenAI key quota.
You can export all your memories at any time from the dashboard. After cancellation, your data is retained for 30 days to allow for recovery or export, then permanently deleted from all systems. We don't hold your data hostage.
Migrate from Zep in a day
EU-hosted. No credit card. Start with a free demo key and your first memory in under 5 minutes.
Get your demo key →Demo key · 1 agent · 100 memories · No expiry