World’s first AI-native database API
Meet ZeroDB
The unified, AI-native serverless database — one API for relational, vector, event, and agent-memory storage.
Cost-optimized vector storage with quantum compression. Free embedding generation included. API-first and MCP-compatible for agent orchestration.
Or skip signup entirely — run npx zerodb-cli init for a working database in seconds, no account or card required. The CLI prints a real claim link — open it any time within 72 hours to attach the project to a persistent account.
https://api.ainative.studio/v1
# Store vector embeddings
curl -X POST https://api.ainative.studio/v1/public/{project_id}/database/vectors/upsert-batch \
-H "X-API-Key: your-api-key" \
-H "Content-Type: application/json" \
-d '{
"vectors": [{
"id": "doc1",
"values": [0.1, 0.2, 0.3],
"metadata": {"text": "AI content"}
}]
}'
# Semantic vector search
curl -X POST https://api.ainative.studio/v1/public/{project_id}/database/vectors/search \
-H "X-API-Key: your-api-key" \
-H "Content-Type: application/json" \
-d '{
"query_vector": [0.1, 0.2, 0.3],
"top_k": 10,
"include_metadata": true
}'What is ZeroDB?
The world's first AI-native serverless database, combining vector storage, relational data, event streaming, and agent memory in one unified API.
Vector Database
HNSW-indexed vector search over high-dimensional embeddings, with free embedding generation built in — no separate vector store or embeddings API to pay for.
Real-time Events
Stream events and build reactive applications with our event store.
Enterprise Security
SOC 2 Type II compliant with end-to-end encryption.
Global Edge
Deploy to 200+ edge locations for sub-10ms latency worldwide.
Free embedding generation
Embeddings, on the house
Every ZeroDB project can generate vector embeddings with the BAAI/bge model family — free, with no external API key, so you stop paying a separate bill just to turn text into vectors.
# Free BAAI/bge model, no external API key needed
curl -X POST https://api.ainative.studio/v1/public/{project_id}/embeddings/generate \
-H "X-API-Key: your-api-key" \
-H "Content-Type: application/json" \
-d '{
"texts": ["ZeroDB is a serverless vector database"],
"model": "bge-base-en-v1.5"
}'
# -> { "embeddings": [[0.0123, -0.0456, ...]] }Why it matters
- No OpenAI or Cohere API key required — generation happens server-side
- Same request can upsert the resulting vector directly into your index
- Included on every plan, including Free — not a paid add-on
Cognitive memory system
ZeroMemory for AI agents
Multi-tier cognitive memory with auto-consolidation, importance-weighted decay scoring, and 100% Recall@1 on the LongMemEval benchmark (ICLR 2025).
Working Memory
Active task context kept in fast-access storage for ongoing reasoning chains.
Episodic Memory
Past interactions and session history with temporal ordering and decay scoring.
Semantic Memory
Long-term knowledge base with importance-weighted consolidation and retrieval.
Six memory API methods
100% Recall@1
LongMemEval benchmark — ICLR 2025
Knowledge graph + GraphRAG
Context Graph
15 graph endpoints for entity extraction, multi-hop traversal, hybrid GraphRAG search, and decision trace logging — all in one unified graph API.
Auto Entity Extraction
3-tier pipeline: Regex → GLiNER → LLM for robust entity recognition at any scale.
Knowledge Graph
Typed entities and relationships with temporal validity and contradiction detection.
GraphRAG
Hybrid vector + graph search combining semantic similarity with graph traversal.
Decision Traces
Agents learn from past reasoning — log, replay, and improve decision chains.
Entity Deduplication
Exact, fuzzy, and semantic matching to keep your knowledge graph clean.
17 Domain Ontologies
Pre-built ontology templates for finance, healthcare, legal, tech, and more.
Learn more15 graph endpoints
IDE native integration
MCP integration
Two MCP servers giving your AI IDE direct access to ZeroDB — from agent memory to full database operations, all from natural language.
zerodb-memory-mcp
npm i ainative-zerodb-memory-mcp
18 tools for agent memory, write-back integrations, and plan artifacts — remember, recall, search, embed, manage context windows, write to Slack/Gmail/Calendar/GitHub/Notion, and persist plans across sessions.
18 tools·Memory focused
zerodb-mcp-server
pip install zerodb-mcp
77 tools for complete database access — vectors, tables, events, files, memory, and graph operations.
77 tools·Full database
Supported IDEs
npx zerodb-cli init6 published packages
Framework integrations
Works with the tools you already use — LangChain, LlamaIndex, MCP servers, or plain REST.
LangChain
Drop-in vector store for LangChain RAG pipelines.
pip install langchain-zerodbLlamaIndex
Native LlamaIndex vector store integration.
pip install llama-index-vector-stores-zerodbZeroDB CLI
Initialize and manage ZeroDB projects from the terminal.
npx zerodb-cli initPython MCP
77-tool MCP server for Python environments.
pip install zerodb-mcpnpm MCP
18-tool memory MCP server for Node.js IDEs.
npm i ainative-zerodb-memory-mcpREST API
Any HTTP client — no SDK required.
curl https://api.ainative.studio/v110 API capabilities
Everything you need to build AI applications
Vector Search
Semantic similarity search with HNSW indexing
Knowledge Graph
GraphRAG — search entities and relationships alongside vectors
NoSQL Tables
Flexible schema-less document storage
SQL Queries
Full SQL support with PostgreSQL compatibility
Event Streaming
Real-time event store and subscriptions
Agent Memory
Persistent memory for AI agents
File Storage
S3-compatible object storage
Dedicated PostgreSQL
NewManaged Postgres with pgvector and row-level security
Lakehouse
NewDuckDB SQL analytics over your data — no ETL pipeline
Quantum Enhanced
NewQNN compression for up to 4x storage efficiency
Performance benchmarks
Built for production scale
Numbers that matter when you're shipping AI products to real users.
Built for AI applications
From RAG pipelines to multi-agent swarms — ZeroDB handles every AI data pattern.
RAG Applications
Document Q&A with semantic search and vector retrieval. Combine full-text and vector hybrid search for maximum recall.
AI Agent Memory
Persistent multi-tier memory across sessions. Working, episodic, and semantic memory with auto-consolidation.
Multi-Agent Swarms
Shared or isolated memory coordination across agent networks. Namespaced access with full audit trails.
Knowledge Graphs
Entity extraction and multi-hop graph queries. Build dynamic knowledge bases that evolve with your data.
Recommendation Systems
Behavior tracking and semantic similarity scoring. Power personalized recommendations with vector embeddings.
Semantic Search
Full-text and vector hybrid search in one API call. No need to manage separate search and vector systems.
How we compare
One API covers everything below — vectors, graph, Postgres, NoSQL, files, events, SQL analytics, agent memory, and RLHF. Most alternatives cover one or two rows and leave the rest to a separate vendor.
| Feature | ZeroDB | Databricks | Pinecone | Mem0 | Zep / Graphiti |
|---|---|---|---|---|---|
| Core data | |||||
| Vector search | |||||
| Dedicated PostgreSQL | Lakebase | — | — | — | |
| NoSQL tables | Delta tables | — | — | — | |
| File storage | — | — | — | ||
| Event streaming | — | — | — | ||
| Lakehouse SQL analytics | — | — | — | ||
| Quantum compression | — | — | — | — | |
| AI & agent features | |||||
| Free embeddings | 150 free DBUs/mo | — | — | — | |
| Agent memory (multi-tier) | Partial | — | |||
| Knowledge graph | Partial | — | Basic | ||
| GraphRAG | Partial | — | — | — | |
| Decision traces | — | — | — | — | |
| Domain ontologies | 17 | — | — | — | — |
| Auto-enrichment hooks | Partial | — | — | — | |
| RLHF data collection | — | — | — | ||
| Plan Artifacts | — | — | — | ||
| Developer experience | |||||
| MCP server | 77 tools | — | — | ||
| Serverless | Cloud | Self-host | |||
“Partial” means the competitor has a real but narrower or less-integrated version of that capability (e.g. Databricks’ knowledge-graph support requires a separate Neo4j integration rather than a native graph store). Comparison based on publicly available documentation as of November 2026.
Simple pricing
All plans include free embedding generation — saving you thousands in OpenAI API costs.
Save 50-80% vs alternatives. Pinecone + OpenAI embeddings typically costs $200+/mo. Supabase + pgvector starts at $75+/mo. ZeroDB includes embeddings FREE.
Works with the tools you already use
Developer testimonials
“ZeroDB replaced 4 different services in our stack. Vector search, document storage, event streaming, and agent memory — all in one API. Setup took 10 minutes.”
Senior AI Engineer
Y Combinator startup
“The MCP server integration is a game changer. My agents in Cursor can query, write, and search our knowledge base directly from the chat window.”
Founding Engineer
AI dev tools startup
“We migrated from Pinecone + OpenAI embeddings and cut our infrastructure bill by 70%. ZeroDB's free embeddings alone paid for the switch.”
CTO
Series A AI company
Developer resources
Everything you need to ship faster.
API Docs
Complete reference for all endpoints.
Getting Started
Up and running in under 15 minutes.
GitHub
Open source SDKs and examples.
Changelog
New features and improvements.
Community
Discord for developers and builders.
Blog
Tutorials, guides, and use cases.
Also building on the AINative platform? Explore ainative.studio for LLM routing, agent orchestration, and the Echo developer program.
Ready to build the future?
Deploy your first ZeroDB instance in under 60 seconds
Free for developers. Scales when you do.
Launch ZeroDB