Milestone: Gemini completed Transaction #000001 through MarginGlow AI Signal
Today we completed MarginGlow’s first external agent-to-agent intelligence transaction.
A Gemini 3.7 Flash agent connected to MarginGlow through our live remote MCP server, selected the analyse_small_business_opportunities tool and requested specialist business intelligence.
MarginGlow AI Signal then:
- Analysed the public business website.
- Returned six evidence items.
- Produced three ranked commercial opportunities.
- Recorded the completed interaction in our authenticated Signal ledger.
- Returned the stored result safely when Gemini replayed the request with the same idempotency key.
The completed external transaction is now permanently recorded as Transaction #000001.
Our internal dashboard separates genuine external activity from MarginGlow quality assurance. Internal tests never consume an external transaction number, allowing judges to distinguish development activity from real agent-to-agent use.
The architecture combines Gemini, Google Cloud Run, Firestore, Streamable HTTP MCP and a structured Signal format designed for use by general-purpose AI agents.
This demonstrates the direction behind MarginGlow AI Signal: a general-purpose agent can retain its relationship with the user while calling MarginGlow for specialist, evidence-based small-business intelligence.
The first transaction used a complimentary Signal allowance. Paid agent-to-agent Signals are the next commercial milestone.
Explore MarginGlow AI Signal: https://www.marginglow.com/signal
Log in or sign up for Devpost to join the conversation.