# ZeroDB — Full Reference (llms-full.txt) > Generated: 2026-11-20 > This is the comprehensive, single-file reference for ZeroDB, per the llms.txt > convention (https://llmstxt.org/). It combines the full API surface, the > authentication flow, and every real SDK quickstart into one file for agents that > want the complete picture without following links. For a shorter summary, see > https://zerodb.ainative.studio/llms.txt. For a single-language reference, see > https://zerodb.ainative.studio/llms/{python,javascript,go}.txt. --- ## Part 1: API Surface, Authentication & Core Capabilities (This section mirrors the root llms.txt verbatim — see that file for the canonical, actively-maintained version.) # ZeroDB > The Agentic Database — serverless, with vector search, agent memory, knowledge graphs, and free embeddings ## When to Use ZeroDB Reach for ZeroDB when your task is one of: - **Giving an agent persistent memory across sessions** — store facts, retrieve them by semantic similarity, let memories decay in relevance over time, forget on request (GDPR). Use the `remember`/`recall`/`forget` tools or `POST /api/v1/public/zerodb/{project_id}/database/memory/batch/upsert`. - **Semantic/vector search over your own documents or embeddings** — HNSW-indexed vector search with free BAAI/bge embedding generation built in, so you don't need a separate embeddings API or vector store. Use `POST /api/v1/public/zerodb/{project_id}/embeddings/search`. - **Building a knowledge graph or GraphRAG pipeline** — automatic entity extraction, typed relationships, hybrid vector+graph retrieval, decision traces for auditability. - **An AI agent needs its own database with no human in the loop** — call `POST /api/v1/public/instant-db` with no auth and `{"agree_terms": true}` to provision a temporary project + API key in one request, no signup form, no credit card. The key and project work immediately and expire in 72 hours unless claimed under a real account. - **You need NoSQL tables, event streaming, file storage, or a dedicated Postgres instance** alongside the above, in one API instead of stitching together separate services. Don't reach for ZeroDB for: a general-purpose relational-only workload with no vector/memory/graph component (a plain managed Postgres is simpler), or a conversational message-exchange protocol between agents (ZeroDB's MCP tools are read-only discovery/data tools, not a chat transport). ## Fuller References - [Comprehensive single-file reference](https://zerodb.ainative.studio/llms-full.txt) (API + auth + every SDK quickstart) - Per-SDK references: [Python](https://zerodb.ainative.studio/llms/python.txt), [JavaScript](https://zerodb.ainative.studio/llms/javascript.txt), [Go](https://zerodb.ainative.studio/llms/go.txt) ## API Base URL https://api.ainative.studio/v1 ## Authentication Direct developer/API-key use: Bearer token via Authorization header or X-API-Key header. Agent-delegated (OAuth 2.1, scoped, standards-based): discover via [authorization-server metadata](https://api.ainative.studio/.well-known/oauth-authorization-server) (RFC 8414) and [protected-resource metadata](https://api.ainative.studio/.well-known/oauth-protected-resource) (RFC 9728), then exchange credentials at `POST https://api.ainative.studio/api/v1/cloud/auth/token` (client_credentials grant; scopes include zerodb.read, zerodb.write, zerodb.admin, memory.recall, memory.store, mcp.connect, mcp.execute — full list at `GET https://api.ainative.studio/api/v1/cloud/auth/scopes`). An agent acting on behalf of a human can use the identity-claim flow at `POST https://api.ainative.studio/agent/identity` and `POST https://api.ainative.studio/agent/claim`. Full guide: [Agent Authentication](https://zerodb.ainative.studio/docs/agent-authentication) (machine-readable: [auth.md](https://ainative.studio/auth.md)). Note: issued scopes are not yet enforced on ZeroDB's own data-plane routes — see the guide for current limitations. ## Core Capabilities ### Vector Search - HNSW indexing with free BAAI/bge embeddings - Semantic similarity search with metadata filtering - Batch upsert and search operations ### Agent Memory (ZeroMemory) - remember: Store facts with auto-embedding and entity extraction - recall: Semantic search with importance-weighted decay scoring - forget: GDPR-compliant deletion by ID, user, session, or time range - reflect: Agent self-reflection on stored memories - profile: Build user profiles from memories - relate: Entity relationship queries ### Context Graph - Automatic entity extraction (3-tier: regex, GLiNER, LLM) - Knowledge graph with typed relationships and temporal validity - GraphRAG: hybrid vector + graph search - Entity merge, deduplication, and timeline - Configurable domain ontologies (17 templates) - Decision traces for agent reasoning auditability ### Additional - NoSQL tables with JSON document storage - Event streaming for agent workflows - S3-compatible file storage - Dedicated PostgreSQL instances - 97-tool MCP server ## Quick Start `npx zerodb-cli init` Guided walkthrough (first vector, first memory, first graph query): [Quickstart](https://zerodb.ainative.studio/quickstart) ## Product Pages Dedicated references for each capability above: - [Agent Memory (ZeroMemory)](https://zerodb.ainative.studio/zeromemory) — remember/recall/forget/reflect, decay scoring, GDPR deletion - [GraphRAG & Context Graph](https://zerodb.ainative.studio/graphrag) — entity extraction, typed relationships, hybrid vector + graph retrieval - [ZeroDB Functions](https://zerodb.ainative.studio/functions) — server-side hooks that run on data change, no separate service to deploy - [Lakehouse](https://zerodb.ainative.studio/lakehouse) — query your vectors, tables and events as Parquet without a separate warehouse - [RLHF](https://zerodb.ainative.studio/rlhf) — collect agent feedback (1-5 ratings, interactions, error traces) to fine-tune on your own data - [ZeroDB Local](https://zerodb.ainative.studio/zerodb-local) — run the full stack on your own machine for development and air-gapped deployments - [Showcase](https://zerodb.ainative.studio/showcase) — real applications built on ZeroDB - [Open-source showcase](https://zerodb.ainative.studio/showcase/open-source) - [Blog](https://zerodb.ainative.studio/blog) — deep dives, benchmarks, and guides; per-section index: https://zerodb.ainative.studio/blog/llms.txt ## Comparisons How ZeroDB differs from adjacent tools, with feature-level detail: - [vs Letta](https://zerodb.ainative.studio/compare/letta) (agent memory frameworks) - [vs Mem0](https://zerodb.ainative.studio/compare/mem0) (agent memory layer) - [vs Pinecone](https://zerodb.ainative.studio/compare/pinecone) (managed vector database) - [vs Weaviate](https://zerodb.ainative.studio/compare/weaviate) (open-source vector database) ## Company - [About](https://zerodb.ainative.studio/about) - [Contact](https://zerodb.ainative.studio/contact) - [Privacy Policy](https://zerodb.ainative.studio/privacy) ## MCP Servers Remote (no install, no credentials required): - [Streamable HTTP endpoint](https://zerodb.ainative.studio/.well-known/mcp) Protocol revision 2025-06-18. Anonymous initialize, tools/list, tools/call, resources/list and resources/read for product and API discovery. Local (stdio, for access to your own project data with your API key): - zerodb-memory (18 tools — 9 memory + 5 write-back + 4 plan artifacts): `npx -y ainative-zerodb-memory-mcp` [Source + AGENTS.md + plugin.json](https://github.com/AINative-Studio/ainative-zerodb-memory-mcp) - zerodb-full (97 tools): `npx -y ainative-zerodb-mcp-server` [Source](https://github.com/AINative-Studio/ainative-zerodb-mcp-server) ## For AI Coding Agents Working in this codebase or integrating with ZeroDB from your own repo? Start with [AGENTS.md](https://github.com/AINative-Studio/ainative-zerodb-memory-mcp/blob/main/AGENTS.md) for conventions, test commands, and how this MCP server is structured. ## Documentation [API Documentation](https://api.ainative.studio/api/v1/docs) Integration guides: - [Agent Authentication](https://zerodb.ainative.studio/docs/agent-authentication) (OAuth 2.1, scopes, identity claim) - [MCP setup](https://zerodb.ainative.studio/docs/mcp-setup) - [LangChain](https://zerodb.ainative.studio/docs/langchain) - [LlamaIndex](https://zerodb.ainative.studio/docs/llamaindex) - [React SDK](https://zerodb.ainative.studio/docs/react-sdk) - [Next.js SDK](https://zerodb.ainative.studio/docs/nextjs-sdk) - [Python SDK](https://zerodb.ainative.studio/docs/python-sdk) ## Pricing Hobbyist $5/mo, Pro $49/mo, Scale $149/mo, Enterprise Custom. Source of truth: [Pricing](https://zerodb.ainative.studio/pricing) (generated from src/config/pricing.ts — see core#274/#276 if this ever looks stale again). --- ## Part 2: Per-SDK Quickstarts Each subsection below mirrors the corresponding public/llms/{lang}.txt file — see those files individually for a lighter-weight, single-language reference. ### Python (public/llms/python.txt) # ZeroDB — Python > Python reference for ZeroDB: the agentic database, with vector search, agent memory, > knowledge graphs, and free embeddings. This file is the Python-specific companion to > https://zerodb.ainative.studio/llms.txt — see that file for the full API surface and > https://zerodb.ainative.studio/llms-full.txt for the comprehensive single-file reference. ## Packages - `pip install zerodb-mcp` — REST client SDK for Python environments (not an MCP protocol server, despite the name — see "MCP server" below for the real one). - `pip install zerodb-cli` — CLI: `zerodb-cli init`, `zerodb-cli status`. - `pip install langchain-zerodb` — drop-in `VectorStore` for LangChain RAG pipelines. - `pip install llama-index-vector-stores-zerodb` — native LlamaIndex vector store integration. ## Authentication Two methods, either works on every authenticated request: 1. `X-API-Key: ` header (preferred) 2. `Authorization: Bearer ` header Get a key instantly with zero signup (72-hour claimable session): ```python import requests resp = requests.post("https://api.ainative.studio/api/v1/public/instant-db", json={}) creds = resp.json() # -> {"api_key": "...", "project_id": "...", "postgres_url": "...", "mcp_config": {...}, "expires_at": "..."} ``` Or register at https://zerodb.ainative.studio to get a permanent key. Agent-delegated OAuth 2.1 (scoped, standards-based) is also supported — see https://zerodb.ainative.studio/docs/agent-authentication for the full flow. ## Quickstart: vector search ```python import requests API_KEY = "your-api-key" PROJECT_ID = "your-project-id" BASE = f"https://api.ainative.studio/api/v1/projects/{PROJECT_ID}" headers = {"X-API-Key": API_KEY} # Generate an embedding (free BAAI/bge model, no external API key needed) embedding = requests.post( f"{BASE}/embeddings/generate", headers=headers, json={"texts": ["ZeroDB is a serverless vector database"], "model": "bge-base-en-v1.5"}, ).json()["embeddings"][0] # Store the vector requests.post( f"{BASE}/database/vectors/upsert", headers=headers, json={ "namespace": "docs", "id": "doc-1", "vector_embedding": embedding, "metadata": {"source": "docs", "text": "ZeroDB is a serverless vector database"}, }, ) # Search results = requests.post( f"{BASE}/database/vectors/search", headers=headers, json={"namespace": "docs", "vector_embedding": embedding, "top_k": 5}, ).json() ``` ## Quickstart: agent memory (ZeroMemory) ```python import requests headers = {"X-API-Key": "your-api-key"} base = "https://api.ainative.studio/api/v1/public/memory/v2" # Store requests.post(f"{base}/remember", headers=headers, json={ "session_id": "my-agent", "content": "User prefers dark mode", "metadata": {"type": "preference"}, }) # Recall results = requests.post(f"{base}/recall", headers=headers, json={ "session_id": "my-agent", "query": "What are the user's preferences?", "top_k": 5, }).json() ``` ## LangChain integration ```python from langchain_zerodb import ZeroDBVectorStore vectorstore = ZeroDBVectorStore( api_key="your-api-key", project_id="your-project-id", namespace="langchain-docs", ) vectorstore.add_texts(["ZeroDB is the agentic database"]) results = vectorstore.similarity_search("database for AI agents", k=5) ``` ## LlamaIndex integration ```python from llama_index_vector_stores_zerodb import ZeroDBVectorStore vector_store = ZeroDBVectorStore( api_key="your-api-key", project_id="your-project-id", namespace="llamaindex-docs", ) ``` ## MCP server (ainative-zerodb-mcp-server) The real MCP protocol server is distributed on npm, not PyPI — `pip install zerodb-mcp` is the REST client SDK above, not this server: ```bash npx -y ainative-zerodb-mcp-server # or use the IDE auto-config CLI, which configures either server for you: npx zerodb-cli init ``` 97 tools covering vector search, NoSQL tables, file storage, events, memory, graph, Postgres, embeddings, RLHF feedback, and quantum-inspired search. Hosted (no install) alternative: https://mcp.ainative.studio/zerodb-full (see https://zerodb.ainative.studio/.well-known/mcp.json for the live, authoritative server listing). ## API surface summary See https://zerodb.ainative.studio/llms.txt for the full endpoint list (vectors, memory, graph, NoSQL tables, events, file storage, Postgres provisioning) and https://zerodb.ainative.studio/llms-full.txt for the comprehensive reference with every endpoint, request/response shape, and all SDK quickstarts in one file. ## Documentation - API Reference (Swagger UI): https://api.ainative.studio/docs - PyPI: https://pypi.org/project/zerodb-mcp/ - Source: https://github.com/AINative-Studio/zerodb-python-sdk - Website: https://zerodb.ainative.studio --- ### JavaScript / TypeScript (public/llms/javascript.txt) # ZeroDB — JavaScript / TypeScript > JavaScript/Node.js reference for ZeroDB: the agentic database, with vector search, > agent memory, knowledge graphs, and free embeddings. This file is the JS/TS-specific > companion to https://zerodb.ainative.studio/llms.txt — see that file for the full API > surface and https://zerodb.ainative.studio/llms-full.txt for the comprehensive single-file > reference. ## Packages - `npm install ainative-zerodb-memory-mcp` — 18-tool MCP server for agent memory, built for Node.js-based IDEs/agents (Claude Code, Cursor, etc). - `npx zerodb-cli init` — one-command CLI that auto-configures the MCP server in your IDE config (Claude Code, Cursor, Windsurf, VS Code). No full-surface JS/TS REST SDK is published yet — use plain `fetch`/`axios` against the REST API below, or the npm MCP server for memory operations from an MCP-capable agent. ## Authentication Two methods, either works on every authenticated request: 1. `X-API-Key: ` header (preferred) 2. `Authorization: Bearer ` header Get a key instantly with zero signup (72-hour claimable session): ```javascript const resp = await fetch("https://api.ainative.studio/api/v1/public/instant-db", { method: "POST", headers: { "Content-Type": "application/json" }, body: JSON.stringify({}), }); const creds = await resp.json(); // -> { api_key, project_id, postgres_url, mcp_config, expires_at } ``` Or register at https://zerodb.ainative.studio to get a permanent key. Agent-delegated OAuth 2.1 (scoped, standards-based) is also supported — see https://zerodb.ainative.studio/docs/agent-authentication for the full flow. ## Quickstart: vector search ```javascript const API_KEY = "your-api-key"; const PROJECT_ID = "your-project-id"; const BASE = `https://api.ainative.studio/api/v1/projects/${PROJECT_ID}`; const headers = { "X-API-Key": API_KEY, "Content-Type": "application/json" }; // Generate an embedding (free BAAI/bge model, no external API key needed) const embedRes = await fetch(`${BASE}/embeddings/generate`, { method: "POST", headers, body: JSON.stringify({ texts: ["ZeroDB is a serverless vector database"], model: "bge-base-en-v1.5" }), }); const { embeddings } = await embedRes.json(); const embedding = embeddings[0]; // Store the vector await fetch(`${BASE}/database/vectors/upsert`, { method: "POST", headers, body: JSON.stringify({ namespace: "docs", id: "doc-1", vector_embedding: embedding, metadata: { source: "docs", text: "ZeroDB is a serverless vector database" }, }), }); // Search const searchRes = await fetch(`${BASE}/database/vectors/search`, { method: "POST", headers, body: JSON.stringify({ namespace: "docs", vector_embedding: embedding, top_k: 5 }), }); const results = await searchRes.json(); ``` ## Quickstart: agent memory (ZeroMemory) ```javascript const headers = { "X-API-Key": "your-api-key", "Content-Type": "application/json" }; const base = "https://api.ainative.studio/api/v1/public/memory/v2"; // Store await fetch(`${base}/remember`, { method: "POST", headers, body: JSON.stringify({ session_id: "my-agent", content: "User prefers dark mode", metadata: { type: "preference" }, }), }); // Recall const recallRes = await fetch(`${base}/recall`, { method: "POST", headers, body: JSON.stringify({ session_id: "my-agent", query: "What are the user's preferences?", top_k: 5 }), }); const results = await recallRes.json(); ``` ## MCP server (ainative-zerodb-memory-mcp) ```bash npm install ainative-zerodb-memory-mcp # or auto-configure your IDE's MCP settings directly: npx zerodb-cli init ``` 18 tools across three groups: - Memory (9): `zerodb_store_memory`, `zerodb_search_memory`, `zerodb_get_context`, `zerodb_embed_text`, `zerodb_semantic_search`, `zerodb_clear_session`, `zerodb_synthesize_context`, `zerodb_configure_auto_context`, `zerodb_get_auto_context_config`. - Write-back (5): `zerodb_slack_send`, `zerodb_gmail_reply`, `zerodb_calendar_create`, `zerodb_github_create_issue`, `zerodb_notion_create_page`. - Plan artifacts (4): `zerodb_plan_create`, `zerodb_plan_get`, `zerodb_plan_update`, `zerodb_plan_history`. For the full 97-tool surface (vectors, NoSQL tables, files, events, Postgres), run the other MCP server instead (`npx -y ainative-zerodb-mcp-server`) — see https://zerodb.ainative.studio/llms/python.txt. ## API surface summary See https://zerodb.ainative.studio/llms.txt for the full endpoint list (vectors, memory, graph, NoSQL tables, events, file storage, Postgres provisioning) and https://zerodb.ainative.studio/llms-full.txt for the comprehensive reference with every endpoint, request/response shape, and all SDK quickstarts in one file. ## Documentation - API Reference (Swagger UI): https://api.ainative.studio/docs - npm: https://www.npmjs.com/package/ainative-zerodb-memory-mcp - Website: https://zerodb.ainative.studio --- ### Go (public/llms/go.txt) # ZeroDB — Go > Go reference for ZeroDB: the agentic database, with vector search, agent memory, > knowledge graphs, and free embeddings. This file is the Go-specific companion to > https://zerodb.ainative.studio/llms.txt — see that file for the full API surface and > https://zerodb.ainative.studio/llms-full.txt for the comprehensive single-file reference. ## Package ```bash go get github.com/AINative-Studio/zerodb-go-sdk ``` ## Authentication Two methods, either works on every authenticated request: 1. `X-API-Key: ` header (preferred) 2. `Authorization: Bearer ` header Get a key instantly with zero signup (72-hour claimable session): ```go resp, err := http.Post( "https://api.ainative.studio/api/v1/public/instant-db", "application/json", bytes.NewBufferString("{}"), ) // resp.Body -> {"api_key": "...", "project_id": "...", "postgres_url": "...", "mcp_config": {...}, "expires_at": "..."} ``` Or register at https://zerodb.ainative.studio to get a permanent key. Agent-delegated OAuth 2.1 (scoped, standards-based) is also supported — see https://zerodb.ainative.studio/docs/agent-authentication for the full flow. ## Quickstart: vector search ```go package main import ( "bytes" "encoding/json" "fmt" "net/http" ) const ( apiKey = "your-api-key" projectID = "your-project-id" ) func base() string { return fmt.Sprintf("https://api.ainative.studio/api/v1/projects/%s", projectID) } func doPost(path string, payload any) map[string]any { body, _ := json.Marshal(payload) req, _ := http.NewRequest("POST", base()+path, bytes.NewBuffer(body)) req.Header.Set("X-API-Key", apiKey) req.Header.Set("Content-Type", "application/json") resp, err := http.DefaultClient.Do(req) if err != nil { panic(err) } defer resp.Body.Close() var out map[string]any json.NewDecoder(resp.Body).Decode(&out) return out } func main() { // Generate an embedding (free BAAI/bge model, no external API key needed) embedResp := doPost("/embeddings/generate", map[string]any{ "texts": []string{"ZeroDB is a serverless vector database"}, "model": "bge-base-en-v1.5", }) embeddings := embedResp["embeddings"].([]any) embedding := embeddings[0] // Store the vector doPost("/database/vectors/upsert", map[string]any{ "namespace": "docs", "id": "doc-1", "vector_embedding": embedding, "metadata": map[string]string{"source": "docs", "text": "ZeroDB is a serverless vector database"}, }) // Search results := doPost("/database/vectors/search", map[string]any{ "namespace": "docs", "vector_embedding": embedding, "top_k": 5, }) fmt.Println(results) } ``` ## Quickstart: agent memory (ZeroMemory) ```go const memBase = "https://api.ainative.studio/api/v1/public/memory/v2" // Store doPostAbs(memBase+"/remember", map[string]any{ "session_id": "my-agent", "content": "User prefers dark mode", "metadata": map[string]string{"type": "preference"}, }) // Recall results := doPostAbs(memBase+"/recall", map[string]any{ "session_id": "my-agent", "query": "What are the user's preferences?", "top_k": 5, }) ``` ## API surface summary See https://zerodb.ainative.studio/llms.txt for the full endpoint list (vectors, memory, graph, NoSQL tables, events, file storage, Postgres provisioning) and https://zerodb.ainative.studio/llms-full.txt for the comprehensive reference with every endpoint, request/response shape, and all SDK quickstarts in one file. Note: the Go module currently wraps the REST API surface directly; there is no separate Go MCP server — both real MCP servers are npm packages: `npx -y ainative-zerodb-mcp-server` (97 tools, full surface) or `npx ainative-zerodb-memory-mcp` (18 tools, agent memory) from an MCP-capable agent/IDE instead. ## Documentation - API Reference (Swagger UI): https://api.ainative.studio/docs - Source: https://github.com/AINative-Studio/Go-SDK - Website: https://zerodb.ainative.studio --- ## Documentation - API Reference (Swagger UI): https://api.ainative.studio/docs - Getting Started: https://zerodb.ainative.studio/docs - Agent Authentication Guide: https://zerodb.ainative.studio/docs/agent-authentication - Website: https://zerodb.ainative.studio - Summary (this file's shorter companion): https://zerodb.ainative.studio/llms.txt - Per-SDK exports: https://zerodb.ainative.studio/llms/python.txt, /llms/javascript.txt, /llms/go.txt