Works with the AI Agents you already use
Token Cost Estimator
Project LLM API costs and token usage by analyzing prompts, datasets, and traffic volumes.
$1
Token Cost Estimator
Example session with this skill installed
Estimate the cost for 10k daily requests using gpt-4o. System prompt is prompts/system.md, average user input is 800 tokens, and expected output is 400 tokens. Use prices.json.
- Read your context and instructions
- Compiled the token cost estimator
I have processed the token counts for your system prompt and calculated the projected costs based on your traffic volume of 10,000 daily requests.
LLM Cost Projection: GPT-4o
| Metric | Value |
|---|---|
| System Prompt (prompts/system.md) | 1,250 tokens |
| Avg. User Input | 800 tokens |
| Avg. Model Output | 400 tokens |
| Total Tokens per Request | 2,450 tokens |
Financial Breakdown (USD)
| Period | Input Cost | Output Cost | Total Spend |
|---|---|---|---|
| Per Request | $0.01025 | $0.00600 | $0.01625 |
| Daily (10k req) | $102.50 | $60.00 | $162.50 |
| Monthly (30 days) | $3,075.00 | $1,800.00 | $4,875.00 |
Context & Cache Optimization
- Context Usage: 2,450 / 128,000 (1.9%) - Low risk of overflow.
Cache Potential: 51% of your input is the static system prompt. Enabling prompt caching could reduce daily input costs by up to $25.60.
Next steps
- Review
prices.jsonto ensure the $5.00/$15.00 per 1M tokens rate is current. - Run a batch count on
conversations.jsonlto see if real user inputs vary from the 800-token estimate. - Compare these figures against a
gpt-4o-miniprice profile to see potential savings.
Connects securely to your tools. The creator never sees your data.
About this skill
Predicting LLM expenses shouldn't be a guessing game. This skill eliminates the manual math and spreadsheet errors when budgeting for AI features. It provides technical estimates for token counts and total API spend based on real files and traffic projections.
What it does
- Token counting calculates volumes across prompts, documents, datasets, and codebases.
- Cost projection estimates daily and monthly spend using custom traffic volumes.
- Context auditing flags requests that risk exceeding specific model context windows.
- Model comparison side-by-sides different providers and pricing tiers for the same workload.
- Caching analysis models the financial impact of prompt caching and batch discounts.
How it works
- Source analysis where the skill counts tokens in your local
.md,.jsonl, or.csvfiles. - Pricing configuration where you provide a JSON file with current per-million-token rates.
- Parameter input including expected output lengths and daily request volume.
- Report generation showing the breakdown of input, output, and cached costs.
Frameworks & tools
The skill utilizes Python scripts for processing. It supports the tiktoken library for cl100k_base accuracy or uses a calibrated heuristic for prose, code, and CJK text.
Why this beats prompting it yourself
General LLMs often hallucinate token counts and fail at complex multi-step financial arithmetic. This tool uses structured scripts and real file parsing to ensure math remains consistent and verifiable across your entire dataset.
Use cases
- Estimate the monthly burn for a new RAG feature before implementation.
- Compare GPT-4o vs Claude 3.5 Sonnet costs for a specific document processing pipeline.
- Audit a
.jsonlfine-tuning dataset to check for context window violations. - Calculate potential savings from implementing prompt caching on static system instructions.
Known limitations
Does not ship with built-in prices; requires a prices.json file. Heuristic counts are estimates and vary by model unless tiktoken is installed and specified.
Requires a prices.json file for cost logic. Heuristic counts are approximate. Accuracy depends on providing current provider rates.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 month ago