# CostCompass

> Spend-intelligence dashboard for solo developers. CostCompass pulls usage from twelve AI and compute providers (Anthropic, OpenAI, Google, DeepSeek, OpenRouter, ElevenLabs, GitHub, Cloudflare, Vercel, RunPod, PostHog, QuantConnect) into one month-to-date total, a forecast, and a per-provider breakdown. Provider API keys stay in a browser-side encrypted vault and are not stored on the server.

## Guides

- [AI Coding Costs](https://costcompass.cc/guides/cost/): Where the money actually goes when you code with AI — most of it is the context an agent reads, not the code it writes. Guides on the real cost drivers.
- [AI Cost Tracking Across Providers](https://costcompass.cc/guides/): Your AI spend is scattered across a dozen dashboards, each in its own units. Multi-provider cost tracking rolls it into one running total and forecast.
- [AI Cost Tracking Tools](https://costcompass.cc/guides/tracking/): The real ways to track what you spend on AI — provider dashboards, spreadsheets, observability tools, and billing dashboards — and what each one costs you in effort.
- [AI Cost Tracking Tools: The 3 Ways They Get Your Numbers](https://costcompass.cc/guides/tracking/ai-cost-tracking-tools/): AI cost tracking tools get your spend three ways: routing traffic, instrumenting code, or reading the bill. How they compare, and where CostCompass fits.
- [CostCompass vs. a Spreadsheet for AI Cost Tracking](https://costcompass.cc/guides/tracking/costcompass-vs-spreadsheet/): CostCompass vs. a spreadsheet for AI cost tracking — where by-hand roll-up goes stale, and what a one-click dashboard with forecasts does instead.
- [How to Enable Prompt Caching: A Step-by-Step Guide](https://costcompass.cc/guides/caching/enable-prompt-caching/): Enable prompt caching in five steps — structure a stable prefix, turn it on per provider, confirm the cache-read counter, and measure the net saving.
- [How to Forecast Your AI Spending Before the Bill Lands](https://costcompass.cc/guides/spend/forecast-ai-spending/): Forecast your AI spending before the bill lands — CostCompass turns seven days of usage into a next-month estimate across every provider.
- [How to Read Claude's Cache Usage Dashboard](https://costcompass.cc/guides/caching/claude-cache-dashboard/): Read Claude's Console Caching page to tell whether prompt caching is working — no logging required: did it read, did it pay, and which request segment broke.
- [How to Track AI Costs Across Multiple Providers](https://costcompass.cc/guides/tracking/track-ai-costs/): Four ways to track what you're spending on AI across every provider — and how to keep one running total and a forecast without changing any code.
- [How to Track Anthropic (Claude) API Costs](https://costcompass.cc/guides/claude-cost-tracking/): How Anthropic bills Claude by model and token, and how to track usage, forecast next month, and keep your key encrypted in your browser.
- [How to Track Atlas Cloud API Costs](https://costcompass.cc/guides/atlas-cloud-cost-tracking/): How Atlas Cloud bills a prepaid, multi-modal account per model, and how CostCompass folds in a Coding Plan for one running total and forecast.
- [How to Track Cloudflare Workers, Workers AI, R2 & Workflows Costs](https://costcompass.cc/guides/cloudflare-cost-tracking/): Cloudflare bills Workers in requests, Workers AI in Neurons, R2 in storage and operations, and Workflows in steps. Price it into one running cost and a forecast.
- [How to Track DeepSeek API Costs](https://costcompass.cc/guides/deepseek-cost-tracking/): DeepSeek API cost tracking — how its prepaid, balance-only billing works, and how CostCompass turns the draining balance into a running total and forecast.
- [How to Track ElevenLabs API Costs](https://costcompass.cc/guides/elevenlabs-cost-tracking/): ElevenLabs bills voice work in characters and credits, not money. Here's how its billing works — and how to track the real cost with a live forecast.
- [How to Track GitHub Actions, Codespaces & Copilot Costs](https://costcompass.cc/guides/github-cost-tracking/): Track GitHub Actions, Codespaces, and Copilot costs — how the minute, core-hour, and AI-credit metering works, turned into a running cost and forecast.
- [How to Track Google Cloud & Gemini Costs](https://costcompass.cc/guides/google-cost-tracking/): How Google bills your whole Cloud account — Gemini, Vertex AI, Compute, Storage — and how to track usage, exact billed cost, and a forecast in one view.
- [How to Track OpenAI API Costs](https://costcompass.cc/guides/openai-cost-tracking/): Track your OpenAI API costs — read org-wide spend with an Admin key, break it down by model, and forecast next month, key encrypted in your browser.
- [How to Track OpenRouter API Costs](https://costcompass.cc/guides/openrouter-cost-tracking/): How OpenRouter bills one prepaid balance across many model vendors, and how CostCompass breaks that spend out by model with a running total and forecast.
- [How to Track PostHog Costs](https://costcompass.cc/guides/posthog-cost-tracking/): PostHog bills analytics by the event and replay by the recording, past each free tier. Here's how pricing works and how to track real cost with a forecast.
- [How to Track RunPod GPU, Serverless & Volume Costs](https://costcompass.cc/guides/runpod-cost-tracking/): How RunPod bills GPU pods and serverless by the second, volumes by the gigabyte, against a prepaid balance — and how to track the cost with a live forecast.
- [How to Track Sentry Costs](https://costcompass.cc/guides/sentry-cost-tracking/): Sentry meters errors and spans but keeps the cost in its own console. CostCompass prices that usage into a running total and forecast across your whole stack.
- [How to Track xAI (Grok) API Costs](https://costcompass.cc/guides/xai-cost-tracking/): xAI (Grok) cost tracking — how xAI reports the exact cost it billed, which credential reads it, and how CostCompass turns it into a running total.
- [LLM Cost Management: How to Control What Your Models Cost](https://costcompass.cc/guides/spend/llm-cost-management/): LLM cost management is a loop: see which model drives cost, pull the right lever, then check the bill moved. Here are the levers and how to read them.
- [Managing AI Spend](https://costcompass.cc/guides/spend/): Knowing this month's total is the start. Managing AI spend means forecasting where it's heading and controlling what your models cost before the bill lands.
- [Prompt Caching](https://costcompass.cc/guides/caching/): Prompt caching can cut model costs sharply — or quietly add to them. Guides on how it works, how to turn it on, how to read the savings, and how to fix it when it doesn't.
- [Prompt Caching Explained: When It Saves Money and When It Costs More](https://costcompass.cc/guides/caching/prompt-caching-explained/): Prompt caching is a bet on reusing a stable prefix within the cache TTL, not an on/off switch — learn when it saves money, when it costs more, and how to tell.
- [The Hidden Economics of AI Coding](https://costcompass.cc/guides/cost/hidden-economics-ai-coding/): AI coding costs come from more than generated code. See how context, prompt caching, model routing, gateways, and observability shape the real bill.
- [What an AI API Cost Dashboard Is — and How to Get One](https://costcompass.cc/guides/tracking/ai-api-cost-dashboard/): An AI API cost dashboard prices every provider's usage into one money figure and forecasts the month. What one shows, and the two ways it gets its numbers.
- [What the OpenClaw Ban Taught Us About the Real Cost of AI](https://costcompass.cc/guides/caching/openclaw-caching-cost-lesson/): Anthropic cut off subscription use of OpenClaw and blamed prompt caching — and the explanation holds up. Your cache hit rate is your real token price.
- [Why Cutting Generated Code Barely Lowers Your AI Bill](https://costcompass.cc/guides/cost/ai-coding-token-costs/): AI coding agents spend most of their tokens reading and re-reading your repo, not writing code. See where the tokens really go — and what lowers the bill.
- [Why Your Prompt Cache Isn't Working: A Troubleshooting Guide](https://costcompass.cc/guides/caching/prompt-cache-troubleshooting/): Prompt caching fails silently — the request still succeeds. Diagnose a stuck cache with three counters: did it write, did it read, and did it actually save?

## Blog

- [You're Probably Over-Counting Your AI Agent's Context Cost](https://costcompass.cc/blog/posts/measuring-agent-context-cost/): Access count × file size over-counts an AI agent's duplicate-load token cost by ~13×. The honest unit is ingested tokens, measured from session transcripts.
- [Why "Hello" Costs Thousands of Tokens](https://costcompass.cc/blog/posts/why-hello-can-cost-thousands-of-ai-tokens/): A one-word prompt to an AI coding agent can hide thousands of tokens: repository exploration, tool output, build logs, and accumulated session state.
- [Ponytail Is Asking the Right Question](https://costcompass.cc/blog/posts/ponytail-right-question/): Ponytail cut the code an AI wrote by half, but the bill by only a fifth. That gap — and the cost that lands months later — is the part worth talking about.
- [The Bill That Didn't Match](https://costcompass.cc/blog/posts/the-bill-that-didnt-match/): My app said I'd spent $12 on Gemini. The bill said $32. Chasing that gap down, and never being blindsided by a provider bill again, is why CostCompass exists.
- [Introducing the CostCompass Almanac](https://costcompass.cc/blog/posts/introducing-the-costcompass-almanac/): A field journal for what your AI and cloud stack actually costs — the numbers, the surprises, and why we're keeping ours out in the open.

## About

- [CostCompass](https://costcompass.cc/): What it is, the dashboard, and the vault/broker model
- [Security](https://costcompass.cc/security/): Encryption architecture and how keys are handled
- [Terms](https://costcompass.cc/tos/): Terms of service
- [Privacy](https://costcompass.cc/privacy/): Privacy policy