Frequently Asked Questions

Kronvex is a persistent memory API for AI agents. Instead of building your own vector database, embedding pipeline, and similarity search from scratch, you call three endpoints:

• POST /remember — store a memory with an automatic embedding
• POST /recall — semantic search with confidence scoring
• POST /inject-context — retrieve a formatted memory block ready to inject into an LLM prompt

Memories persist across sessions. Every agent gets its own isolated namespace.

Click Get Demo Key on any page. Enter your name, email, and a quick description of what you're building. Your key arrives instantly — no credit card, no waiting.

Your key starts with kv- and is displayed once on screen and sent by email. Store it securely — it grants full API access.

An Agent ID is a UUID that namespaces all memories for a single agent instance. When you call POST /api/v1/agents, Kronvex returns an agent ID. All /remember, /recall, and /inject-context calls use that agent ID in the URL path.

On the Demo plan you get 1 agent. Paid plans allow multiple agents — each with their own isolated memory store.

Yes. The Demo plan is permanently free with no time limit:

• 1 agent
• 100 memories
• Full API access (all three endpoints)
• No credit card required

It's designed for prototyping and evaluation. When you're ready to scale, upgrade to Builder (€29/mo) or higher from your Dashboard.

Kronvex is a REST API — it works with any language. Official SDKs are available for:

• Python — pip install kronvex (sync + async with httpx)
• Node.js / TypeScript — npm install kronvex (ESM + CJS)

Native integrations exist for LangChain, LangGraph, CrewAI, n8n, Dify, Flowise, and AutoGen. See the Integrations page for details.

The confidence score combines three signals:

• Semantic similarity × 0.6 — cosine similarity between query and memory embedding
• Recency × 0.2 — sigmoid decay with a 30-day inflection point (older memories score lower)
• Frequency × 0.2 — log-scaled count of how often a memory has been accessed

The result is a float between 0 and 1. You can use it to filter or rank results on your side.

Kronvex uses OpenAI text-embedding-3-small with 1536 dimensions. Embeddings are generated server-side — you send raw text, we handle the embedding.

Vectors are stored in PostgreSQL via the pgvector extension and indexed with HNSW for sub-10ms p99 recall latency on the hosted infrastructure.

/inject-context is a higher-level endpoint that returns a formatted string you can paste directly into a system prompt. It runs a recall internally and formats the top-N memories as a numbered list with metadata.

Use /recall when you need raw memory objects (e.g., to display in a UI or filter programmatically). Use /inject-context when you want to augment an LLM prompt with minimal code.

The /remember endpoint returns a 429 Quota exceeded error. Existing memories and /recall//inject-context calls continue to work normally.

To resolve this, either upgrade your plan from the Dashboard or delete old memories using DELETE /api/v1/agents/{id}/memories/{memory_id}.

The production base URL is https://api.kronvex.io. All API calls require the header X-API-Key: kv-your-key.

The interactive Swagger docs are available at api.kronvex.io/docs ↗.

Yes. Public endpoints (demo key creation, contact) are rate-limited to prevent abuse. Authenticated API endpoints have generous per-key rate limits suited to production agent workloads.

If you're hitting limits at scale, contact us — Enterprise plans have custom limits.

Five plans cover every stage:

• Free €0 · 1 agent · 100 memories
• Builder €29/mo · 5 agents · 20k memories
• Startup €99/mo · 15 agents · 75k memories
• Business €349/mo · 50 agents · 500k memories
• Enterprise Custom · unlimited agents · unlimited memories · SLA

See the full Pricing page for feature details.

Yes. Upgrades take effect immediately with prorated billing for the remainder of the month. Downgrades take effect at the next renewal date.

Manage everything from the Dashboard → Billing tab.

Go to Dashboard → Billing → Cancel subscription. Your paid plan stays active until the end of the current billing period, then reverts to Demo (1 agent, 100 memories).

Data retention: if you exceed the new plan's memory limit (whether cancelling or downgrading to a lower plan), excess memories are scheduled for permanent deletion 10 days after the plan change. You have that window to export or delete what you need before the purge runs.

All major credit and debit cards (Visa, Mastercard, Amex) via Stripe. Bank transfers and invoicing are available on Enterprise plans — contact us to arrange.

All data is stored in the European Union — Supabase Frankfurt (AWS eu-central-1). The API itself runs on Railway (EU region). No data leaves the EU.

Kronvex is EU-hosted by conviction, not just by compliance checkbox.

Yes. Key GDPR provisions covered:

• Data stored exclusively in the EU (Frankfurt)
• Right to erasure — delete agents or individual memories via API or Dashboard
• Data portability — export your memories via API
• No data sold to or shared with third parties

Read our full Privacy Policy and Security page for details.

API keys are SHA-256 hashed before storage. The plaintext key is shown to you exactly once (on creation) and never stored. Even Kronvex cannot recover your key — if lost, generate a new one from the Dashboard.

Yes, at any granularity:

• Delete a single memory: DELETE /api/v1/agents/{id}/memories/{memory_id}
• Delete an agent and all its memories: DELETE /api/v1/agents/{id}
• Close your account entirely: contact hello@kronvex.io

Install via pip:

pip install kronvex — sync client
pip install "kronvex[async]" — async client (httpx)

Then:

from kronvex import KronvexClient

client = KronvexClient(api_key="kv-your-key")
client.remember(agent_id, "User prefers dark mode")

Use the KronvexMemory class as a drop-in for LangChain's memory interface. It stores and retrieves memories via the Kronvex API on each chain run.

Full integration guide: docs → LangChain integration. You can also see live examples on the Integrations page.

Yes. Kronvex exposes an MCP (Model Context Protocol) server that lets Claude, Cursor, and other MCP-compatible tools call remember, recall, and inject_context as tools natively.

Setup guide: docs → MCP server.