In Mem0
Your memory is unstructured — chat-history snippets, user facts as free text.
Comparison
nlqdb vs Mem0
Pick Mem0 if you want an opinionated memory primitive — add / search / forget — tuned for LLM agent contexts. Pick nlqdb if your agent also needs to query structured data, run reports, and migrate its own schema.
The same goal, two ways.
> users who logged in this week and viewed pricing
In your HTML
<nlq-data goal="users who logged in this week and viewed pricing"></nlq-data>The structured behavioural slice an agent extracts is nlqdb's lane; Mem0's lane is the unstructured fact recall.
What's different
Four dimensions that decide it
| Dimension | nlqdb | Mem0 | Note |
|---|---|---|---|
| Structured rows + typed columns | Mem0 stores facts as text + vectors; nlqdb stores typed rows in Postgres. | ||
| Natural-language queries | |||
| Vector search over chat history | |||
| MCP server |
Show 5 more rows
| Dimension | nlqdb | Mem0 | Note |
|---|---|---|---|
| Auto-migration via NL ('add a `priority` field') | |||
| Multi-agent / multi-device shared tenant | |||
| Explicit forget / TTL semantics | |||
| Aggregations + reporting queries | |||
| Open source |
shipped · partial · not shipped
When to choose nlqdb
- Your agent stores structured rows ('user X bought Y on Z') the agent later queries.
- You want one MCP server that does provisioning, memory, and reporting.
- Multiple agents on multiple devices share one tenant-scoped database.
- The schema needs to evolve as the agent learns ('add a `priority` field').
When to choose Mem0
- Your memory is unstructured — chat-history snippets, user facts as free text.
- Vector recall over fuzzy strings matters more than typed SQL.
- You want a memory-only primitive; the agent is wired into another data layer.
- A managed memory tier with explicit forget semantics is on your shortlist.
Questions buyers ask
- Can I use Mem0 for fuzzy facts and nlqdb for structured data?
- Yes — they're complementary. Mem0 handles 'remember the user prefers Celsius', nlqdb handles 'list the user's orders this month'. Both can sit behind one MCP-aware agent; nlqdb's MCP server exposes `nlqdb_query`, which materialises Postgres on first reference, so the structured side is self-provisioned.
- Is nlqdb a vector database?
- No. nlqdb is Postgres-first (ClickHouse for analytical engines in Phase 2). For vector recall over unstructured strings, Mem0, Pinecone, or pgvector are the right shape.
- How does my agent provision an nlqdb database autonomously?
- The MCP server exposes `nlqdb_query` — your agent calls it with a goal in English, the server materialises Postgres + schema on first reference, and returns the answer bound to the agent's tenant. There's no separate create-DB verb to call first.
- Does nlqdb support forget / TTL like Mem0?
- Anonymous-mode databases auto-sweep after 72h; authenticated tables don't ship TTL semantics yet. If forget is core to your agent's memory model, Mem0 fits better today.
- Can multiple agents share the same nlqdb database?
- Yes — tenant-scoped `sk_live_*` keys give each agent access to the same data. Per-device tagging is supported via `sk_mcp_*` keys minted with `(mcp_host, device_id)` claims, so the dashboard shows 'Cursor on macbook-air ran 14 queries today'.
Try nlqdb in 30 seconds
No sign-in. The anonymous database lasts 72 hours; adopt it with one click if you keep it.
Start with a goal →