Agent inference cost reconciler

Show developers exactly where self-hosting agents is cheaper or pricier than SaaS

Problem

One user asked how to get inference for personal projects, complaining that major providers keep cutting subscription quotas; another built an AI agent running-cost calculator but only for a single call; a developer shared that their monthly API bill for building their own tool is painful. These threads show there is no trustworthy tool for self-host vs hosted cost comparison.

Solution

A local-first cost reconciliation desktop tool: import your agent call logs or manually enter model, token counts, call volume, and automatically compare self-hosted (including GPU depreciation and electricity) against managed services (OpenAI, Anthropic, AWS Bedrock) on a monthly basis; generate charts over projects and timelines and suggest the break-even call volume for switching.

Users

Indie developers running side projects or internal tools on LLM APIs with monthly bills above a few tens of dollars.

MVP

Manual input of model, call volume, per-token price; built-in price table for major hosted and self-hosted options; monthly cost comparison chart with break-even point; CSV import for call logs; fully local, no data upload.

Revenue

One-time desktop purchase, or a free base tier with paid multi-project management and historical price tracking.

Why now

Inference quotas are shrinking across the board, forcing developers to redo the math, and existing calculators only cover single calls, missing the full self-host vs hosted comparison.