Skip to main content

Analyser — Token Counter & Cost Estimation Tool

Access: All users (always available, no admin configuration needed)

The Analyser is a utility tool for counting tokens and estimating costs across multiple LLM models. Use it to understand how much of a model's context window your text will consume and what it will cost — before you send it to an agent.

Analyser results showing token counts and costs — Replace with actual screenshot​

What is the Analyser?​

When you work with AI agents, every piece of text (prompts, responses, system instructions) is broken down into tokens — small chunks of text that the model processes. The number of tokens directly affects:

  • Cost — You're charged per token by the LLM provider
  • Context window usage — Every model has a maximum number of tokens it can process at once
  • Performance — Longer prompts mean slower responses

The Analyser lets you paste any text and instantly see:

  • How many tokens it contains
  • How much it would cost across different models
  • How much of each model's context window it uses

Why it matters: Before deploying a prompt to production or sending a long document to an agent, you want to know:

  • Will it fit in the model's context window?
  • How much will each request cost?
  • Which model gives the best cost/performance ratio?

Demo Video​

How to Use the Analyser​

Step 1: Open the Analyser​

Click Analyser in the sidebar. It's always visible to all users — no admin configuration needed.

Step 2: Enter Your Text​

Paste or type your text in the input area. This can be:

  • A system prompt you're writing
  • A sample user message
  • A document you plan to send to an agent
  • Any text you want to analyze

Step 3: View the Results​

The tool displays results in real-time as you type — no need to click a button.

What you'll see:

MetricDescription
Character CountTotal number of characters in your text
Token CountsNumber of tokens for each supported model, displayed simultaneously
Estimated CostCost per model based on current pricing for input tokens

The results are shown for multiple Google models simultaneously (Gemini Pro, Gemini Flash, etc.), organized by vendor. This lets you compare at a glance.

Understanding the Results​

Token Count​

  • Different models tokenize text differently — the same text may result in slightly different token counts across models.
  • A rough rule of thumb: 1 token ≈ 4 characters in English (but this varies by language and model).

Cost Estimation​

  • The cost shown is for input tokens only (since you're analyzing text before sending it).
  • Actual cost per request also includes output tokens (the agent's response), which vary based on the response length.
  • The pricing is based on the current published rates for each model.

Context Window Usage​

  • If your text is close to a model's context window limit, you'll have very little room left for the conversation.
  • Recommendation: Your system prompt should ideally use no more than 10–20% of the model's context window, leaving plenty of room for user messages and agent responses.

Common Use Cases​

Use CaseWhat to Do
Prompt EngineeringWrite your system prompt in the Prompts module, then paste it here to check token usage across models. Ensure it fits within the context window with room to spare.
Cost PlanningBefore deploying an agent, paste a representative prompt to estimate per-request costs. Multiply by expected daily volume for a cost projection.
Model ComparisonPaste the same text and compare token counts and costs across models. Choose the most cost-effective model for your use case.
Document Length CheckBefore sending a large document to an agent, paste it here to verify it fits within the model's context window.

Tips for Beginners​

  • Use it before the Prompts module — Draft your prompt, check the token count here, then refine if it's too long.
  • Check costs before going to production — A prompt that costs $0.001 per request may seem cheap, but at 10,000 requests/day, that's $10/day just for the system prompt.
  • Shorter is usually better — If two prompts produce similar agent quality but one uses fewer tokens, go with the shorter one. It's cheaper and leaves more room for context.
  • Pair with Prompts and Cost Control — Use the Analyser for pre-deployment estimation, the Prompts module for management, and Cost Control for monitoring actual production costs.

Tip: The Analyser currently supports Google models for token counting. For cross-provider cost comparison, use the token count as a rough estimate for other providers (counts are typically within 10% across major LLMs for English text).