Native JSON-RPC 2.0 Engine, Dynamic Tool Registries & SSE Real-Time Streaming
Deploy AuthHub as your primary MCP Server. Connect Anthropic Claude Desktop, Cursor, Zed, or autonomous agent fleets directly via HTTP SSE. AuthHub delivers spec-compliant JSON-RPC 2.0 lifecycle management, adaptive tool discovery based on caller identity, cryptographic Merkle audit anchoring, and real-time tool push notifications.
Full compliance with official MCP specifications across all standard methods, error codes, and transport mechanisms.
Supports MCP spec versions 2026-07-28, 2025-11-25, 2025-06-18, 2025-03-26, and 2024-11-05. Echoes client version or negotiates gracefully during initialize.
Dynamically generates tools/list based on the authenticated user's relationship graph in SpiceDB. Agents only see tools they are authorized to call, eliminating prompt window clutter and token waste.
Supports persistent HTTP streaming via GET /mcp with Accept: text/event-stream. Delivers server notifications, list change events, and execution progress in real time.
Serves contextual documentation, database schemas, and clinical guidelines via resources/list and resources/read. Supports dynamic URI parameter templates.
Provides curated prompt templates via prompts/list and prompts/get. Standardizes clinical reasoning steps, database query constraints, and multi-step workflows.
Every tool execution is hashed, anchored into a cryptographic Merkle tree in CockroachDB, and mirrored to immutable MinIO S3 object storage for 7-year regulatory compliance.
| Method | Direction | Description | Standard Response |
|---|---|---|---|
| initialize | Client → Server | Negotiates protocol version and advertises server capabilities | result: { protocolVersion, capabilities, serverInfo } |
| notifications/initialized | Client → Server | Signals initialization completion. Idempotent acceptance. | HTTP 202 Accepted (No response body) |
| ping | Client → Server | Liveness probe. Supported pre-authentication. | result: {} |
| tools/list | Client → Server | Returns dynamic catalog of available tools with input schemas | result: { tools: [...] } |
| tools/call | Client → Server | Executes a specific tool with validated arguments | result: { content: [...], isError: false } |
| resources/list | Client → Server | Returns list of readable context resources | result: { resources: [...] } |
| resources/read | Client → Server | Reads contents of a specific resource URI | result: { contents: [...] } |
| prompts/list | Client → Server | Returns governed prompt templates | result: { prompts: [...] } |
| prompts/get | Client → Server | Renders prompt messages with supplied arguments | result: { messages: [...] } |
| notifications/tools/list_changed | Server → Client (SSE) | Server push notification when tool permissions mutate | SSE event: message \n method: ... |
Select your development environment or AI client for ready-to-use configuration files.
claude_desktop_config.json:{
"mcpServers": {
"authhub": {
"command": "npx",
"args": [
"-y",
"@authhub/mcp-remote",
"--url", "https://api.authhub.cloud/mcp",
"--token", "YOUR_BEARER_TOKEN"
]
}
}
}Launch tools with full audit guarantees, adaptive schemas, and real-time streaming today.