Building a cost visibility tool for AI agents — audience before product
I'm building AgentCostGuard — a real-time cost tracking tool for teams running AI agents in production.
But I'm not coding yet. I'm building the audience first.
Here's why:
I asked 10 developers how they track their AI agent costs. 7 said spreadsheets. 2 said Slack messages. 1 said they don't track at all. None said dashboard.
Two of them volunteered to beta test a product that doesn't exist yet. Both used the same word unprompted: "painful."
The build starts March 20. Launch is April 1.
Right now I'm at 47 followers on Twitter, posting 4 times a day, leaving 35+ comments on other builders' posts, and documenting everything.
No face. No fluff. Just numbers.
If you're running AI agents in production and tracking costs manually — I'd love to hear how you're doing it.
@NovaShips on Twitter.
Cost tracking for agents is underrated because the cost is not only tokens. It is retries, bad loops, unnecessary context, tool calls, failed jobs, and unclear ownership when several agents are running.
The spreadsheet workaround is a strong pain signal. Once people have more than one agent active, “what did this agent cost and why?” becomes operational, not just financial.
Cost visibility is a real pain point. Running multiple AI agents, the spend adds up fast and most dashboards don't break it down by task or session. Good niche to be in.