Build a hosted MCP server (Claude + ChatGPT connector) with semantic search and admin panel

We need to build a hosted MCP server (Claude + ChatGPT connector) with semantic search and admin panel for our agency. ABOUT THE PROJECT We are a video strategy and production company serving B2B SaaS marketing teams. We're turning our video planning methodology (a few hundred documents plus a large library of analyzed video examples) into a hosted MCP server, so clients can plan videos from their own AI assistant (Claude and ChatGPT) and our internal team can use a deeper tier of the same system. IP protection is the point: content stays on our server, clients get answers, and access is revocable per client. WHAT YOU'D BUILD - A hosted MCP server (Streamable HTTP, always on) exposing tools for retrieval, video type recommendations, script and storyboard drafting support, brief filling, and brief submission. - Semantic search / vector retrieval over 200-300 documents plus hundreds of structured video records - Two access tiers (client vs internal), per-client credentials, revocable access, usage logging - An admin tool I can operate without a developer: content re-sync from Google Docs, batch video ingestion, client management - Integrations: Google Docs, ClickUp, and email for request intake. HOW WE'LL WORK - Independent freelancers only, no agencies. I want to work one-on-one with the person actually writing the code, from the first call to launch. - Four milestones: MVP (internal testing), private beta (per-client auth, recording transcription), personalization (per-client profiles, reliability hardening), commercial launch (billing, dashboard, rate limits). We commit milestone by milestone, starting with the MVP. - I'm a non-technical founder. I need plain-language explanations, options with tradeoffs, and predictable weekly updates. - You work independently; I test everything personally between milestones. - NDA and IP assignment required. A detailed brief goes to shortlisted candidates. YOUR PROPOSAL MUST INCLUDE 1. Rough per-milestone pricing, with the video-analysis pipeline and transcription broken out as line items. 2. Recommended stack and architecture, with reasoning a non-developer can follow. 3. One-time build cost vs ongoing monthly costs (hosting, vector database, APIs). 4. Timeline per milestone and your weekly availability. 5. Your assumptions and what you'd need from me. 6. MCP servers or AI integrations you've built (links or descriptions) Proposals that skip per-milestone pricing won't be considered.