Obsidian is brilliant for your own thinking. But when your AI agent needs memory, a local markdown vault isn't a memory system — it's a workaround that breaks at scale.
An accurate look at where each tool excels — and where each breaks down.
| Dimension | Obsidian | Kronvex |
|---|---|---|
| Designed for | Human note-taking & PKM | ✓ AI agent persistent memory |
| Memory retrieval | File-based search / keyword | ✓ Semantic search (pgvector cosine similarity) |
| Cloud access | ✕ Local vault (sync costs extra) | ✓ REST API — works everywhere |
| CI/CD & serverless | ✕ Requires local filesystem | ✓ HTTP endpoint — no filesystem needed |
| Multi-user | ✕ Single-user only | ✓ Unlimited users via isolated agent IDs |
| Multi-agent | ✕ One vault per person | ✓ Isolated memory namespaces per agent |
| Semantic search | ● Plugin required (Smart Connections) | ✓ Built-in — every recall uses embeddings |
| Memory auto-growth | ✕ Manual note creation | ✓ POST /remember — one API call |
| Recency & frequency scoring | ✕ None — all notes equally weighted | ✓ similarity × 0.6 + recency × 0.2 + freq × 0.2 |
| GDPR / right to erasure | Manual file deletion | ✓ Built-in erasure, TTL, export API |
| Pricing | Free (local) · $10/mo sync · $25/mo Publish | ✓ From €19/mo — all features included |
| Setup time (for agents) | 30+ min (install, vault, MCP plugin, sync) | ✓ 5 minutes to first memory |
| Scale ceiling | 1 developer · 1 device · 1 project | ✓ Unlimited agents & users |
Obsidian is a world-class PKM. These limitations only matter when you try to use it as an agent memory backend.
Four things Kronvex does that no Obsidian vault can match at scale.
Kronvex stores every memory as a 1536-dimension embedding (text-embedding-3-small). Recall queries by cosine similarity — no local process, no personal OpenAI key, no Obsidian running. Ask "what auth pattern did we choose?" and get the relevant memory even if it says "JWT with 15min TTL, refresh is silent."
pgvector · cosine similarity · hosted APIEvery recalled memory is ranked by similarity × 0.6 + recency × 0.2 + frequency × 0.2. Recency uses a sigmoid with a 30-day inflection. Old, unused memories fade gracefully. No manual curation, no vault pruning sessions, no HANDOFF.md overhead.
Kronvex is an HTTP API: /remember, /recall, /inject-context. Works in Docker, Lambda, GitHub Actions, any server, any language. The same API that powers your local workflow can power 10,000 users in production — with zero architecture change.
All data stored in Supabase Frankfurt (eu-central-1). Right to erasure is a first-class API endpoint — not a "find all notes mentioning this email and delete them manually" task. TTL-based memory expiry, data export, and DPA included. Built for European compliance from day one.
Frankfurt · right to erasure · TTL · DPAOne approach requires Obsidian running locally. The other is a plain HTTP call.
# Requires: # 1. Obsidian running on your machine # 2. obsidian-mcp plugin installed # 3. Local vault synced (Obsidian Sync: $10/mo) # 4. Smart Connections for semantic search # (uses your own OpenAI API credits) # Breaks in: Docker · GitHub Actions · any server # Not deployable. Not multi-user. Not a product.
import httpx # Store — works from anywhere httpx.post( "https://api.kronvex.io/api/v1/agents/agent-1/remember", headers={"X-API-Key": "kv-..."}, json={"content": "User prefers dark mode, uses VS Code."} ) # Recall — semantic, cloud-backed ctx = httpx.post( "https://api.kronvex.io/api/v1/agents/agent-1/inject-context", headers={"X-API-Key": "kv-..."}, json={"query": "editor preferences", "max_tokens": 400} ).json()["context"] # Works in Docker, Lambda, CI — no Obsidian needed
These tools solve different problems. Here's the honest breakdown.
Your vault knowledge doesn't disappear — it moves to an API. Three steps.
Create an account at kronvex.io/demo — you get 100 memories, 1 agent, no credit card required. Your kv-... key is ready in seconds.
Obsidian stores everything as .md files. A simple Python script reads each file, splits it into logical chunks (one fact or decision per memory), and POSTs each to /api/v1/agents/{id}/remember. Contact hello@kronvex.io and we'll send you a ready-to-run migration script.
Update your Claude / Cursor / Cline MCP config to point to the Kronvex MCP server instead of the Obsidian plugin. From that point, your agent recalls memories semantically from the API — no vault sync required, works in any environment. Your Obsidian vault stays intact for personal note-taking.
.md files — easy to read programmatically. Export your vault, split each note into logical chunks (one fact or decision per memory), then POST each chunk to POST /api/v1/agents/{id}/remember. Contact hello@kronvex.io for a ready-to-run Python migration script.No credit card. No SDK required. No Obsidian running. Just an API key and three endpoints — works in Docker, Lambda, CI, and every cloud environment.
Get your free API key →