Relay — Meeting-to-Action Agent

What the company appears to be: Relay is a self-reported tool that processes meeting transcripts using GPT-5.6 to extract structured action items, decisions, and open questions. It allows users to approve these items before automatically filing them into Slack, Linear, and Notion.

What changed: The project was built as part of the OpenAI 2026 hackathon. It evolved from an initial approach that used LLM agents for all steps to a more reliable system using deterministic API calls after discovering reliability issues with LLM-based filing.

Single most important open question: Does Relay have any real-world usage or adoption beyond its hackathon prototype? The description provides no evidence of revenue, customers, or traction.

The description states that Relay:

• Processes meeting transcripts
• Uses GPT-5.6 to extract decisions and action items (owner, due date, priority)
• Grounds each extracted item to the exact transcript line it came from
• Requires user approval before filing
• Posts a recap to Slack
• Creates one Linear issue per action item
• Writes a full meeting-notes page in Notion

The product is described as a tool that transforms unstructured meeting data into structured, actionable records across multiple platforms.

The description states:

• The tool is positioned as an automated "Meeting-to-Action Agent"
• It claims to use GPT-5.6 for extraction
• It emphasizes that nothing is filed until the user reviews and approves it
• It describes a shift from agent-based filing to deterministic API calls due to reliability issues

The positioning appears to be evolving from a generic meeting automation tool to one focused on structured data extraction and reliable filing.

Not evidenced. The description does not state who the target customer is or what their specific needs are beyond general meeting follow-through problems.

Not evidenced. The description provides no information about pricing, revenue streams, or business model.

The description states:

• Built with FastAPI + Python backend
• Single HTML frontend file (no build step)
• Uses GPT-5.6 for structured JSON extraction via JSON schema
• Filing to Slack/Linear/Notion uses direct API calls, not LLM in the loop
• Model provider is swappable: OpenAI directly or OpenRouter
• The team tested three different models for filing step, all failed differently
• Fixed reliability issues by switching from agent-based to deterministic API calls

Not evidenced. There is no evidence of revenue, customers, usage metrics, or any traction beyond the hackathon submission.

Not evidenced. The description does not mention competitors or market context.

• The tool was built for a hackathon and has no demonstrated traction
• Reliability issues were discovered during development (three models failed differently)
• The team size is listed as 1 person, suggesting limited resources for scaling
• No evidence of revenue, customers, or adoption beyond the prototype
• The project description lacks any mention of real-world usage or testing

1. What specific meeting problems did you observe that led to building this?
2. Have you tested this with actual teams and users beyond your own?
3. How do you plan to monetize this tool if you intend to commercialize it?
4. What is the timeline for moving from prototype to a production-ready product?
5. Are there any existing tools in this space that you're aware of?