01 / 02
Briefcase Client GPT
Knowledge was fragmented across Asana, email threads, and individual drives. An agency-wide survey I ran found 61% of participants named "centralize client knowledge" as their top pain point. Context got lost. Work quality was inconsistent across departments.
System architecture diagram โ MCP pipeline connecting Asana, Google Drive, and multiple LLM interfaces
Most AI tools at agencies fail for the same reason: they ask people to switch. Briefcase runs on Model Context Protocol โ it lives inside Claude, ChatGPT, and Gemini. Whatever your team already opens in the morning, Briefcase is already there. Supabase vectorizes the knowledge base, Railway hosts the pipeline, and each user gets scoped access so client data never crosses.
Four sources in. One index. Any interface out.
Briefcase exposes 22 MCP tools grouped by verb. Anything an agency team actually does โ retrieving context, triggering a sync, prepping for a call, logging a task โ is a single call away.
Because Briefcase runs on MCP, it doesn't care which LLM you use. Ask the same question in Claude, ChatGPT, or Gemini. Same data. Same cited answer.
Production-deployed on Railway. Rolling out agency-wide with enablement sessions teaching ICs how to plug Briefcase into Webflow, SEMRush, and the ad platforms they already use.
Explore RoccOS.
My portfolio lives inside a fake operating system. Apps, easter eggs, and an AI companion included.
Return to RoccOS