Resources
How We Built an AI Agency Inside Buzz
The complete system behind InferenceSaver's agency infrastructure — from social listening lanes to automated outreach, documented as we built it. Built by Matthew Don.
The Infrastructure
We built a multi-lane signal pipeline that routes social listening data from X (via OpenMagpie's twitter_search connector) into Twenty CRM, through a pipeline server, and into dedicated Buzz channels. Each lane — Automation, Agency, Agency Partners, UGC Creators — has its own channel, its own query set, and its own lane-owner agent that deliberates signals.
The Buzz SDK Wiring
A pipeline server running on node01 routes structured events into Buzz channels via the `/buzz-notify` endpoint, which calls `buzz messages send` directly. This bypasses the relay's broken template interpolation and gives us full control over the message format.
The Go-to-Market Engineer Role
The GTM Engineer funnels people into the product by running the social listening pipeline, hunting for second-source data feeds that give each lane a competitive edge, and measuring performance against real analytics. The role is the engine between signal intake and revenue.
The Product Surfaces
Our agency offers are captured in the repo as structured JSON: Agency (done-for-you builds), Agency Partners (cheaper inference for AI agencies), Automation (end-to-end AI workflow implementation), and Deals (landing page with $10 starter credits, volume discounts, referral program, and DeepSeek V4 Flash at $1/100M tokens).
How You Can Use This
The same systems we built for ourselves are available through our agency. We can build the same infrastructure for your business — websites, AI systems, automations, and the operational layer that connects them. This book documents the playbook.