The Programmable Cloud Built for Autonomous AI Agents
Give Cursor, Claude Desktop, Antigravity, and custom agent workflows safe, authenticated tools to query databases, trigger serverless execution, and manage storage via official Model Context Protocol (MCP).
Ecosystem Compatibility
Works with Your Favorite AI Agent Clients
Cursor AI
Connect directly to Cursor's MCP panel for schema-aware code generation and database migrations.
Claude Desktop
Empower Claude with local stdio MCP tools to analyze live data, write queries, and inspect logs.
Google Antigravity
Native integration for autonomous engineering workflows, automated incident response, and agentic DevOps.
Custom LangChain / LlamaIndex
Use our official Python and TypeScript SDKs to equip your custom autonomous pipelines with cloud capabilities.
Full Cloud Capabilities at the Agent's Fingertips
Instead of forcing agents to write shell scripts or juggle raw credentials, PGB Hub exposes standard, structured JSON-RPC tools with built-in validation and guardrails.
Live Schema & SQL Execution
Agents can inspect live database schema, list tables, and execute parameterized SQL queries safely with automatic schema validation and connection pooling.
Serverless & Compute Dispatch
Trigger pre-warmed serverless microVM runners with sub-second latency, run background worker tasks, and receive real-time execution outputs directly.
Sovereign Object Storage
Store generated images, documents, reports, and embeddings in S3-compatible sovereign object storage with presigned URLs and instant access.
Transactional Mail & Messaging
Allow agents to send verified transactional notifications, dispatch system emails, and publish realtime websocket events to web and mobile subscribers.
Scoped Least-Privilege Keys
Generate dedicated project-scoped API keys restricted strictly to the required database or function, preventing privilege escalation or accidental leakage.
Tamper-Evident Audit Trails
Every tool call executed by an autonomous agent is immutably logged with actor identity, timestamp, input parameters, execution duration, and outcome.
Add PGB Hub to Your Agent in Seconds
PGB Hub's MCP server runs over stdio via npx, seamlessly connecting to your local editor or agent runtime.
{
"mcpServers": {
"pgbhub": {
"command": "npx",
"args": ["-y", "@pgb/mcp"],
"env": {
"PGB_PROJECT_REF": "your-project-id",
"PGB_SECRET_KEY": "pgb_sec_your_scoped_key_here"
}
}
}
}Replace PGB_PROJECT_REF and PGB_SECRET_KEY with your project credentials from the console.
{
"mcpServers": {
"pgb-cloud": {
"command": "npx",
"args": ["-y", "@pgb/mcp"],
"env": {
"PGB_PROJECT_REF": "your-project-id",
"PGB_SECRET_KEY": "pgb_sec_your_scoped_key_here",
"PGB_API_URL": "https://api.pgbhub.com"
}
}
}
}Restart Claude Desktop after updating this file. The hammer icon will show your active PGB tools.
You are an autonomous engineering agent connected to PGB Hub via the Model Context Protocol (@pgb/mcp).
You have safe, authenticated tools to interact with the project:
1. Use 'db_schema' to inspect tables, relations, and columns before writing queries.
2. Use 'db_query' with parameterized SQL statements for reads and mutations.
3. Use 'invoke_function' to test serverless business logic without managing infrastructure.
4. Always verify schema constraints before applying migrations.Zero Prompt or Data Mining
PGB Hub never intercepts, trains on, or logs your queries or embeddings with third parties. Your agent payloads remain strictly confidential.
Fine-Grained Permissions
Restrict agent keys to read-only database queries or specific compute runners. Prevent unwanted table drops or destructive updates.
Sub-Second Tool Latency
Pre-warmed runners and optimized PostgreSQL pooling mean agent tool calls return in milliseconds, keeping LLM chains fast and responsive.
Ready to Supercharge Your AI Agents?
Generate scoped agent credentials in seconds and connect your favorite LLM workflows to sovereign databases, compute, and storage.