The Problem with Open-Source ML Tools
Most open-source tools in the AI ecosystem are marketing vehicles. They demo a feature someone built once, never get maintained, and sit abandoned on GitHub gathering stars.
But some of them are genuinely useful. They solve a specific problem well enough that developers ship them into production.
We built Cost Skills as the latter. Not as a funnel. As actual tools.
What is a Cost Skill?
A Cost Skill is a free, single-prompt Claude Agent Skill that demonstrates one specific feature of Cost.botzone.ai.
Each skill solves a narrow problem. Analyse a single API call. Break down token counts by model. Compare inference costs across providers. Route between Claude 3.5 Haiku and Claude 3.5 Sonnet and see the savings.
One prompt. One answer. No sign-up. No credit card.
The skill is designed to show you what's possible. Then you realise you need this for your entire codebase. Then you sign up for Cost.
Example: The Cost Analyzer Skill
Let me show you the flow.
You paste a Claude API call (the request and response JSON). The Cost Analyzer Skill:
- Parses the request and response
- Extracts token counts (input, output, prompt-cache hits if using Claude's cache)
- Looks up current Claude pricing
- Computes the USD cost
- Shows you the breakdown
Input: 5,000 tokens @ $3 per M = $0.015
Output: 1,200 tokens @ $12 per M = $0.0144
Total: $0.0294
That's it. One request. One answer.
But now you're thinking: I have 2,000 API calls a day in production. How much am I spending? Which routes are burning cash? What if I switched half my traffic to Haiku?
That's when you sign up. Because you need this continuously.
How to Use a Skill
Installation is one command:
npm install @botzone/cost-sdk
Or for Python:
pip install botzone-cost
That's the Cost SDK. It's open-source. MIT license. The SDK wraps your existing Anthropic, OpenAI, or Gemini client:
import Anthropic from "@anthropic-ai/sdk";
import { wrap } from "@botzone/cost-sdk";
const client = wrap(new Anthropic(), {
route: "summariser",
apiKey: process.env.COST_API_KEY, // only needed if you upgrade to the product
});
const message = await client.messages.create({
model: "claude-3-5-sonnet-20241022",
max_tokens: 1024,
messages: [{ role: "user", content: "Your prompt here" }],
});
That's your API call. The SDK intercepts it. Ships token counts to Cost. Zero latency overhead.
If you're using a Cost Skill without the SDK, you just grab the JSON from your API logs and paste it into the Skill prompt. Single analysis. Done.
Why We Built This
Most developer tools get acquired for their distribution, not their code. The code stays proprietary. The tool gets sunset.
We wanted to prove that's backwards. Real tools spread because they're useful. Not because you find them in a sales email.
So we built the SDKs open-source. MIT license. You can fork them. You can modify them. You can vendor them.
The Skills are free prompts that you run in Claude, Copilot, or any Claude-compatible system (Hermes Agent, OpenRouter, etc.).
They're designed for developers who want to see what's happening before they commit to a product. No trust required. Just run the code. See the results.
The Funnel (Written Honestly)
Here's what happens when you use a Cost Skill:
- You paste an API call into the Skill
- You see the cost breakdown
- You realise you have 1,000 of these API calls running daily
- You want to understand total spend, not one-off requests
- You want to see spend by route, by model, by customer
- You want to know which route to downgrade to Haiku without losing quality
- You sign up for Cost
That's not manipulation. That's genuine problem discovery. By the time you're ready for the product, you already know what you're buying. You've seen the signal. You're buying because you need continuous observability, not because a sales pitch convinced you.
Try It
Cost Skills are live. Pick one that matches your immediate problem:
- Cost Analyzer - Break down token costs from a single API call (TypeScript / Python SDK)
- Model Router - Evaluate switching between Claude 3.5 Haiku and Sonnet on a sample of your traffic
- Batch Cost Calculator - Total spend across a production API log
Find them on GitHub at botzone-ai/cost-sdk-ts and botzone-ai/cost-sdk-py. Each repo has the skill examples in the /skills directory.
Run one. See the output. If you need this for production, you know what to do.
If you don't, that's fine. The skill still works. Use it the next time you need a cost breakdown.
That's the deal. Free tool. Genuine signal. No strings.
Head to cost.botzone.ai to sign up when you're ready, or start with the open-source SDK and Skills if you want to keep it local. Either way, you're building with real information.
Start saving today
Know exactly where your LLM money goes.
Cost wraps your Anthropic, OpenAI, and Gemini clients in one line. Free tier covers 100,000 events per month. No card needed.
Start tracking your spend