Feature Request – Context Injection via MCP during Transcript Pipeline

Hi Omi Team,

I’d like to suggest adding MCP tool calls during the transcript processing phase, upstream of the final summary generation.

The Logic: Right now, summaries are based only on the audio transcript. However, many conversations reference external context—events, emails, or project tasks—that are already accessible via connected MCP tools but remain "siloed" during the summarization process.

My Current Implementation: I have already built an MCP server (FastAPI/Python) integrating Microsoft Graph (Calendar, Email, To-Do) into my Omi flow. The value is clear: resolving conversation references against real data results in a "Master Summary" rather than a fragmented one.

Proposed Flow (The "Enrichment Hook"):

1. Entity Extraction: Post-transcript, Omi identifies key entities (dates, people, specific project names).

2. Context Retrieval: Omi triggers relevant MCP "Search" or "List" tools based on those entities.

3. Injected Prompting: The retrieved data is injected into the LLM context window alongside the transcript to generate the final summary.

Example Use Case:

• Transcript: "Let's prep for the Fintech Summit in 5 days."

• Omi Action: MCP queries the calendar for "Fintech Summit," pulls the last 3 related emails from the organizer, and checks for open tasks.

• Enriched Summary: "You discussed the Fintech Summit (Apr 27). Context: Venue is DWTC; you have an outstanding task to 'Confirm AV specs' in MS To-Do."

Why This Matters: This moves Omi from being a passive transcript logger to an Active Contextual Assistant. It bridges the gap between what we say and the data we own.

Thank you