Shipping 1 SaaS per month — Day 2 of Tool #1 (AgentShield)
I'm building 6 SaaS tools in 6 months. Solo. No funding. Real numbers.
Tool #1 is AgentShield — cost observability for AI agents. Track spend per agent in real time, replay every session, kill switch if costs spike.
Why this problem? I talked to 10+ developers running AI agents in production. None of them could tell me what their agents cost per run. The most common tracking method: checking the invoice at the end of the month and hoping it didn't spike.
Where I'm at after 2 days of coding:
- 21 commits, 62 tests passing
- FastAPI backend with 19 database tables
- POST /v1/track endpoint live (the core of the product)
- Auto cost calculation for 30+ AI models
- API key service, rate limiting, plan enforcement
- Next.js frontend initialized (dark mode)
- Python SDK package initialized
Kill criteria: less than $200 MRR at 12 weeks → I kill it publicly and move to Tool #2.
3 beta testers already signed up from Twitter conversations alone. $0 MRR. Launch target: April 1st.
Anyone here running AI agents and dealing with cost visibility? Would love to hear how you're handling it.
