Zdravo AI is a universal AI memory layer that captures and organizes your conversations across ChatGPT, Claude, Gemini, Cursor, Windsurf, and 10+ AI platforms. It uses semantic search to help you find any solution from any chat in under 10 seconds.

How does Zdravo AI work?

Install our Chrome extension or use our bookmarklet to capture conversations with one click. Zdravo AI automatically generates summaries, tags, and embeddings. Use semantic search to find relevant conversations by meaning, not just keywords.

Yes. Zdravo AI uses AES-256 encryption for all stored data. We offer Bring Your Own Key (BYOK) so you can use your own encryption keys. Your data is stored in your own Supabase instance with Row-Level Security.

Which AI platforms does Zdravo AI support?

Zdravo AI supports ChatGPT, Claude, Gemini, Perplexity, Cursor IDE, Windsurf IDE, and 10+ other AI platforms through our Chrome extension and bookmarklet.

What is the difference between free and paid plans?

The free tier includes 15 memories per month, basic search, and browser extension. Pro plans include unlimited memories, advanced search, team features, API access, and priority support. Team plan supports up to 5 users with shared workspaces.

MCP Integration Guide

Zdravo is a memory substrate that any AI agent can read from and write to via MCP. This guide provides copy-paste-ready integration instructions.

Quick Start

# 1. Install the MCP server

npx @zdravoai/mcp-server --api-key YOUR_API_KEY

# 2. Configure your agent (Claude Desktop, Cursor, etc.)

{

"mcpServers": {

"zdravo": {

"command": "npx",

"args": ["@zdravoai/mcp-server", "--api-key", "YOUR_KEY"]

}

Get your API key from Dashboard → Settings → API Keys

Available Tools

Tool

memory.search

Search your memory vault using semantic vector search + keyword matching.

{

"query": "architecture decisions for Q3",

"limit": 5,

"memory_type": "semantic"

}

Tool

memory.create

Save a new memory to your vault. Auto-tags, embeds, and indexes.

{

"content": "Meeting decided to use PostgreSQL for the new service",

"title": "Database Decision - June 2026",

"tags": ["architecture", "database", "decision"]

}

Tool

memory.recall

Recall relevant memories for a given context. Surfaces related knowledge automatically.

{

"context": "Working on authentication refactor",

"limit": 3

}

Tool

auto_inject

Automatically surface relevant memories based on current file context or URL.

{

"file_path": "src/auth/middleware.ts",

"imports": ["express", "jsonwebtoken"]

}

Claude Desktop Setup

Install Claude Desktop from claude.ai/download
Open Claude Desktop → Settings → Developer → Edit Config
Add the Zdravo MCP server configuration to your claude_desktop_config.json:
{
"mcpServers": {
"zdravo": {
"command": "npx",
"args": ["@zdravoai/mcp-server", "--api-key", "sk_zdravo_..."]
}
}
}
Restart Claude Desktop
In a new chat, Claude will automatically have access to memory.search, memory.create, and memory.recall

Cursor Setup

Open Cursor → Settings → MCP Servers
Click "Add MCP Server" and paste:
npx @zdravoai/mcp-server --api-key sk_zdravo_...
Enable the server and verify connection status shows "Connected"
In any chat, Cursor can now search and save memories automatically

API Reference

The MCP server exposes the following endpoints via SSE (Server-Sent Events):

GET /sse — SSE connection endpoint

POST /messages — Tool invocation endpoint

GET /api/mcp/health — Server health check

Troubleshooting

Connection refused

Ensure the API key is valid and starts with sk_zdravo_. Check that the MCP server process is running.

No memories found

Memories are user-scoped. Ensure you're using the correct API key for your account. Run GET /api/mcp/health to verify connection.

Rate limiting

The MCP server enforces 100 requests per minute. If you hit the limit, wait 60 seconds and retry. Enterprise plans have higher limits.

Available Tools

Tool

memory.search

Search your memory vault using semantic vector search + keyword matching.

{

"query": "architecture decisions for Q3",

"limit": 5,

"memory_type": "semantic"

}

Tool

memory.create

Save a new memory to your vault. Auto-tags, embeds, and indexes.

{

"content": "Meeting decided to use PostgreSQL for the new service",

"title": "Database Decision - June 2026",

"tags": ["architecture", "database", "decision"]

}

Tool

memory.recall

Recall relevant memories for a given context. Surfaces related knowledge automatically.

{

"context": "Working on authentication refactor",

"limit": 3

}

Tool

auto_inject

Automatically surface relevant memories based on current file context or URL.

{

"file_path": "src/auth/middleware.ts",

"imports": ["express", "jsonwebtoken"]

}

Claude Desktop Setup

Install Claude Desktop from claude.ai/download
Open Claude Desktop → Settings → Developer → Edit Config
Add the Zdravo MCP server configuration to your claude_desktop_config.json:
{
"mcpServers": {
"zdravo": {
"command": "npx",
"args": ["@zdravoai/mcp-server", "--api-key", "sk_zdravo_..."]
}
}
}
Restart Claude Desktop
In a new chat, Claude will automatically have access to memory.search, memory.create, and memory.recall

Troubleshooting

Connection refused

Ensure the API key is valid and starts with sk_zdravo_. Check that the MCP server process is running.

No memories found

Memories are user-scoped. Ensure you're using the correct API key for your account. Run GET /api/mcp/health to verify connection.

Rate limiting

The MCP server enforces 100 requests per minute. If you hit the limit, wait 60 seconds and retry. Enterprise plans have higher limits.