FamRelay

Product Introduction

1. Definition: FamRelay is a neutral, permissioned relay layer built on the Model Context Protocol (MCP). It is a technical middleware service that enables structured data sharing between disparate AI assistants and large language model (LLM) applications without merging user accounts or exposing chat histories.
2. Core Value Proposition: It exists to solve the problem of AI assistant data silos and context fragmentation within households, small teams, and for individual users of multiple AI tools. Its primary value is enabling selective fact sharing and cross-assistant coordination while maintaining user privacy and control, eliminating the need for manual message forwarding and repetitive explanations.

Main Features

1. Permissioned, Item-Level Sharing: The core technical mechanism is a user-in-the-loop approval workflow. When a user requests a share, their AI assistant (via the MCP connector) extracts structured facts (e.g., date, time, location) from the conversation and presents a proposed share for explicit user approval. Nothing is transmitted until the user confirms. This ensures minimum necessary data transfer and user consent for every action.
2. Neutral MCP-Based Connector: FamRelay operates as an independent service accessible via a standard MCP server endpoint (https://api.famrelay.ai/mcp). This allows it to integrate with any AI assistant or application that supports custom MCP connectors (e.g., Claude Desktop, Cursor, Windsurf, compatible IDEs). This vendor-agnostic architecture prevents lock-in and enables connections between assistants from different providers.
3. Granular Control & Audit Trail: The system provides a visible, queryable record of all shared items. Users can ask their connected assistant for a summary of what was shared, to whom, and the status (delivered, read, expired, revoked). Each share has a configurable expiry date (event-based, custom, or default 30 days) and can be revoked at any time, instructing connected assistants to delete the data.

Problems Solved

1. Pain Point: Context Fragmentation and Manual Handoff Overhead. Critical information (appointments, travel details, school schedules) resides in isolated AI assistant "islands," forcing users to manually re-share information via messages, emails, or repeated conversations, leading to errors and the common refrain, "I thought you knew."
2. Target Audience: Cross-Assistant Households & Teams: Families where members use different AI assistants (e.g., one uses ChatGPT, another uses Claude). Multi-Tool Professionals & Power Users: Individuals who use different AI assistants for different contexts (e.g., Claude for work, ChatGPT for home, Grok in Tesla). Small Teams & Founders: Groups that need to coordinate decisions, deadlines, and client handoffs across members using various AI-augmented tools.
3. Use Cases: Event Coordination: Sharing christening/wedding details (date, time, venue) from a flyer parsed by one assistant to a partner's different assistant. Home Management: Relaying plumber/electrician arrival windows to whoever is home. Travel Planning: Distributing flight itineraries and lodging details to all trip participants. Caregiving Handoffs: Updating family members on appointment schedules or task completion for dependents. Personal Workflow Sync: Passing a decision or deadline noted in a work-focused assistant (Claude in Cursor) to a general-purpose assistant (ChatGPT) used for personal planning.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):

   • Pros:
     • Preserves Privacy: Shares only approved, structured data points, not conversation history or assistant memory.
     • User-Centric Control: Every share requires explicit approval, with clear visibility and revocation capabilities.
     • Platform Agnostic: Works across any AI tool supporting MCP, future-proofing against ecosystem changes.
     • Reduces Cognitive Load: Automates the tedious "messenger" role between people and between one's own tools.
   • Cons & Limitations:
     • Technical Barrier: Requires an assistant that supports custom MCP connectors, excluding many mainstream consumer chatbot interfaces.
     • Honest Revocation Limit: As stated, revoking a share removes future access but cannot retroactively delete information already read by a human or copied elsewhere.
     • Structured Data Focus: Optimized for clear facts (dates, times, tasks). Nuanced, unstructured context or complex instructions may not relay effectively.
     • Network Effect Dependency: Full value for teams/families requires others to also connect their assistants, though solo use provides immediate value.

2. Key Alternatives & Differentiation:

   • Shared Group Chats/Messaging Apps (WhatsApp, Slack): FamRelay differentiates by automating the extraction and structured sharing of specific facts directly into the recipient's AI assistant's context. It moves beyond broadcasting full messages in a chat stream, enabling the AI to proactively use the information in planning and reminders.
   • Single Assistant Ecosystem (e.g., all using ChatGPT Team): FamRelay offers an open-standard alternative to vendor lock-in. It allows a family or team to maintain their preferred, best-of-breed AI tools (Claude for writing, Perplexity for research, Grok in car) while still achieving shared context, unlike a single-provider solution that forces uniformity.
   • Manual Copy-Paste or Screenshot Sharing: FamRelay automates this process with a structured, auditable, and reversible workflow, eliminating the friction and error-proneness of manual transfer.

Frequently Asked Questions (FAQ)

1. Does FamRelay work with ChatGPT or Claude? Yes, FamRelay works with any AI application that can add custom Model Context Protocol (MCP) connectors. This includes Claude Desktop, Cursor, Windsurf, and other MCP-compatible tools. It does not currently integrate directly with web-based chatbots like chat.openai.com, as they do not expose custom MCP functionality. Users must use a compatible desktop/client application.

2. Is my data safe with FamRelay? How is it encrypted? According to its security documentation, FamRelay encrypts every shared item with its own unique key. Access is locked separately for each intended recipient. The service states it does not use shared data to train AI models or sell it. Data is only processed to facilitate the approved relay between the specified user's connected assistants.

3. Can I use FamRelay by myself if no one else I know uses an AI assistant? Absolutely. A primary use case is personal cross-assistant synchronization. By connecting multiple assistants you own (e.g., Claude for work, ChatGPT for home, Grok in your Tesla) to the same FamRelay account, you can approve the relay of facts between them, creating a unified personal context across your tools.

4. What happens when I revoke a shared item? Revoking a share immediately stops future access to that item on FamRelay and sends a instruction to all connected assistants to delete the item from their local context. The key limitation, as stated transparently by FamRelay, is that revocation cannot erase information already read by a human user or copied manually into another system outside of the connected assistants.

5. How much does FamRelay cost? FamRelay is currently free during its early access and pilot phase. The company states it is learning what features provide genuine value before establishing a pricing model. They commit to providing clear communication before any future payment is required.