按开发者管控 AI 编程支出

一款 SaaS，用于跨模型追踪和管控每位开发者的 AI 编程支出，提供预算、告警以及更便宜模型的路由建议。

I want to build a small, bootstrappable product. Help me plan and start the MVP.

Product: Per-developer AI coding spend control
What it is: SaaS that tracks and controls per-developer AI coding spend across models, with budgets, alerts, and cheaper-model routing recommendations.
Who pays: VP of Engineering / platform leads at 200–2,000 person tech companies rolling out coding assistants, paying $1–3k/mo from the FinOps or developer-tools budget
Closest alternatives: CloudZero, Vercel AI Gateway, Portkey, manual GitHub Copilot usage dashboards
The gap: Cloud cost tools don't understand LLM tokens/agents; gateway tools route calls but don't give engineering managers per-person budgets, per-model cost attribution, and nudges to cheaper models. Microsoft had to build this in-house (usage meters, credit caps); everyone else has nothing.
Why now: In the last ~12 months enterprises moved from unlimited AI usage to hard per-employee caps (Microsoft cut limits 10x), so spend governance for coding agents just became mandatory rather than optional.
Moat to build toward: Accumulates cross-org benchmarks of per-developer/per-model spend and ROI that become the reference dataset for AI budget policy; workflow lock-in once budgets and alerts are configured.

可直接粘贴到 Claude Code、Cursor 或任何编程助手的需求说明（英文）。