CLAUDE SKILLS
Your AI assistant, taught to do AI cost math
A Claude Skill is a small folder of instructions your assistant loads exactly when a task matches it. The Weckr plugin ships four of them. Install once and Claude Code knows real model prices, real SDK syntax, and the right way to reason about per user cost and margin. Free, open source, no account needed.
Install in Claude Code
# one time, inside Claude Code
/plugin marketplace add Ghiles3232/weckr-sdks
/plugin install weckr@weckrThat is the whole setup. The four skills below are now available in every session.
How a skill works
You ask a normal question: "which of my plans lose money on AI cost?" No slash command, no special phrasing.
Claude sees the question matches a skill description and loads that skill’s instructions into the conversation, only when relevant.
The answer uses current prices and the shipped SDK syntax, with the math shown so you can challenge any assumption.
The four skills
Wire Weckr into your app
Claude integrates the SDK with the exact two line pattern that ships: the real constructor, the real wk.chat options, the real error types. No invented syntax, no stale examples.
Triggers on: “help me track AI costs per user in my app”
Current LLM prices on tap
Per token prices for OpenAI, Anthropic, Gemini, and Kimi. Claude fetches the live feed at useweckr.com/pricing.json first, so its numbers match what providers charge today, not what they charged at training time.
Triggers on: “what does gpt-5.4-mini cost per million tokens?”
Forecast a feature before you ship
Claude reads your LLM call, estimates tokens, and projects cost per call, per user, and per month against your price point. Always as a range with stated assumptions, never one false precision number.
Triggers on: “how much will this summarizer cost me per user?”
Find the plans that lose money
Give Claude your plans and rough usage. It costs a typical and a heavy user on each plan and flags where the heavy tail goes underwater, the quiet loss flat pricing hides.
Triggers on: “which of my pricing plans lose money on AI cost?”
Why the numbers stay current
Most price tables in docs and prompts go stale the week a provider changes rates. These skills do not, because they sit at the end of a pipeline that starts at the same table Weckr uses to compute real billing.
A weekly watcher also diffs every price table against published provider rates and opens a pull request when something moves, so a price cut like OpenAI’s July 2026 drop on GPT‑5.6 Terra and Luna lands in the feed within days, reviewed by a human before it ships.
Skills or the MCP server?
Both, they answer different questions. The skills give Claude know-how: correct prices, correct syntax, correct method. The Weckr MCP server gives Claude live access to your own dashboard data, so you can ask “which of my users cost me money this month?” and get your real numbers back.