AgentWiki — Dual-Interface Knowledge Platform
A single knowledge base serving both human researchers and autonomous AI agents.
Overview
Knowledge bases fail when humans and AI agents compete for the same interface. I architected AgentWiki as a multi-tenant SaaS with strict RBAC, where humans interact through a React 19 UI while agents authenticate via an MCP server and CLI. Deployed on Cloudflare Workers for edge-native latency, the platform treats agents as first-class users with scoped programmatic access. The architecture separates concerns: the React layer optimizes for human cognition, the MCP layer for machine parsing. Live deployment validates the dual-interface model—one codebase, two interaction modes, zero interface conflict.
Highlights
- 01
Deployed live multi-tenant SaaS with RBAC and agent CLI access
- 02
Built on React 19 and Cloudflare Workers for edge performance
- 03
MCP server enables programmatic agent workflows alongside human UI
System Architecture
Dual-interface architecture: shared RBAC enforces tenant isolation for human and agent paths.
Questions people ask
- How does the platform handle human and AI interactions simultaneously?
- It separates concerns with a React 19 UI for humans and an MCP server for agents, ensuring zero interface conflict.
- What infrastructure supports the application's performance?
- The system is deployed on Cloudflare Workers to leverage edge-native latency and high availability.
- How is access control managed for different user types?
- I implemented strict RBAC and multi-tenancy to securely scope programmatic access for autonomous agents.