Agent Memory API
Persistent memory for AI agents — Mem0-class quality, 1/5th the price
Add facts, recall by semantic similarity, manage per-user memory across sessions. BGE embeddings + sqlite-vss. Multi-tenant by design. $0.00005/event — 5× cheaper than Mem0 Cloud at the same MTEB tier.
How we compare
There is no objective "agent memory quality" benchmark — unlike DER for diarization or F-measure for segmentation. v1.0 calibrates on embeddings quality (MTEB) + feature parity. v1.1 will add LongMemEval and a 1M-event ingest stress test.
| Provider | Quality | Price/event | vs market avg | Position |
|---|---|---|---|---|
| Mem0 Cloud | 9.0/10 | $0.00025 | 57% | — |
| Zep Cloud | 8.8/10 | $0.00050 | 114% | — |
| LangGraph Platform Memory | 8.5/10 | $0.00080 | 183% | — |
| Pinecone (DIY) | 9.2/10 | $0.00020 | 46% | — |
| Brainiall FAST | 8.9/10 | $0.000050(89% cheaper) | 11% | Parity |
Pricing rule: 90% off when inferior · 80% off at parity · 50% off when superior. Position determined by objective benchmark, refreshed quarterly.
Pricing
Discount derived from quality position vs the closest competitor. 90% off when inferior, 80% off at parity, 50% off when superior.
Free
$0/mo
1,000 events/month · fast tier · 1 namespace
Starter
$9/mo
100,000 events/month · 10 namespaces · all formats
Pro
$49/mo
1M events/month · unlimited namespaces · priority queue
Business
$199/mo
10M events/month · dedicated capacity · email + Slack
PAYG: $0.00005/event (BGE-small). HD tier (BGE-large + bge-reranker-large for +1-2pp recall on long-context retrieval) is on the v1.1 roadmap — not yet available.
Three endpoints, multi-tenant by design
# Add a fact (event) to a user's namespace
POST https://api.brainiall.com/v1/memory/{namespace}/add
{"text": "User prefers dark mode UI", "metadata": {"source": "settings"}}
# Semantic search across the namespace
GET https://api.brainiall.com/v1/memory/{namespace}/search?q=ui+preferences&k=5
-> {"results": [{"text": "...", "score": 0.87, "metadata": {...}}, ...]}
# Retrieve a recent timeline (chronological)
GET https://api.brainiall.com/v1/memory/{namespace}/recent?limit=20Why this exists
- Per-user memory for chatbots, AI agents, copilots without rolling your own pgvector + extraction pipeline.
- Multi-tenant: namespace per user, project, or tenant — no cross-leakage by design.
- Open weights: BGE models are MIT-licensed, so we can run them on our own infra and pass the savings through. Mem0 and Zep both pay OpenAI for embeddings.
- HD tier on roadmap: BGE-large + bge-reranker-large for +1-2pp recall on long-context retrieval, coming v1.1.
Press kit & resources
Everything reviewers, integrators and procurement teams typically ask for.
One-page datasheet
Pricing, KPIs and a copy-pasteable curl snippet on one page. Ideal for buyer review.
Download PDFAPI reference
OpenAPI spec, request/response shapes, error codes, rate limits and quota model.
Read docs →More specialty APIs
Same single API key, same usage-based pricing, different problem solved.
