Hi all,
We're evaluating Rovo MCP Server as part of our architecture for an internal knowledge chatbot backed by Confluence. Here's our context:
- Deployment: Standalone web app (not an IDE or desktop client)
- Users: 100–500 employees in a single organisation
- Access pattern: Read-only queries against Confluence
- Query types: Ranging from simple lookups to complex, cross-page synthesis questions
We have a few technical concerns we'd love the community to weigh in on:
Scalability — Is Rovo MCP designed to handle concurrent requests from many users hitting a shared backend, or is it architected primarily for single-user, interactive sessions?
Query complexity — For questions that require reasoning across many pages simultaneously (e.g., aggregating information across an entire Confluence space), does the MCP tool-calling model hold up? Or does it struggle compared to a pre-indexed vector search approach?
Auth model — For a shared app, is a single API token (service account) a supported and stable auth pattern, or is OAuth 2.1 per user the only production-grade option?
Honest best-fit guidance — We don't want to over-engineer or misuse the tool. What types of workflows and applications is Rovo MCP genuinely built and optimised for? We'd rather pivot our architecture now than hit a wall later.
Would really value input from anyone who has shipped something similar or has deep knowledge of Rovo MCP's internals. Thanks!
