

Package your content once as an encrypted, signed payload. Share a handle. Let the receiver's AI render it however they prefer — in Claude, Gemini, ChatGPT, or Manus.
Write your content in PromptShare. It is encrypted with AES-256-GCM, signed, and stored with a unique UUID handle.
Send the 8-word instruction or the handle URL. The receiver pastes it into their AI assistant.
The AI fetches the payload, verifies the signature, asks for rendering preferences, and presents the content.
Content is locked at creation. Receiving AI assistants can only change how it is presented, never what it says.
Every payload is signed with HMAC-SHA256. Recipients can verify the content has not been tampered with.
Automatic scanning for prompt injection patterns before storage. Harmful instructions are blocked at the gate.
Works with Claude, Gemini, ChatGPT, and Manus. One handle, any AI assistant.
Generate API keys for programmatic access. Integrate with your AI workflows via the MCP REST API.
Exposes a standard MCP tool manifest so any MCP-enabled assistant can discover and use the hub natively.
External AI assistants can fetch any public payload with a single GET request. No authentication required for reads.
GET /api/mcp/payloads/:handle
# Response
{
"schema": "prompt-hub/v1",
"handle": "550e8400-e29b-41d4-a716-...",
"title": "Q1 Strategy Brief",
"sealedContent": "...",
"signatureValid": true,
"instructions": {
"step1": "Ask reader: format, tone, detail level.",
"step2": "Confirm style, then render sealedContent.",
"rule": "DO NOT modify sealedContent."
}
}