> ## Documentation Index
> Fetch the complete documentation index at: https://lightsage.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# How Lightsage credits work and how to estimate usage
> Learn how Lightsage credits work, what each platform type costs per run, and how to estimate your monthly usage before choosing a plan.
Credits are the billing unit Lightsage uses for running AI evaluations. Every time Lightsage runs a prompt on an AI platform or tests an API endpoint with a coding agent, it consumes credits. The number of credits a run costs depends on the type of platform — lighter answer engine lookups cost less than full coding agent executions that write and run real code.
## Credit costs by platform type
Different platform types cost different amounts because they require different levels of compute and time to run.
| Platform type | Examples | Credits per run |
| ----------------------------- | --------------------------------------------------------------------------------------------------- | --------------- |
| **AI Search / Answer Engine** | ChatGPT, Perplexity, Google AI Overviews, Gemini Search, Claude, Google AI Mode, Microsoft Copilot | 1 credit |
| **Coding Agent** | Claude Code, OpenAI Codex, Cursor, GitHub Copilot, Gemini CLI, OpenClaw, OpenCode, Hermes, Pi, Kilo | 5 credits |
| **Model (API Performance)** | DeepSeek, Kimi, Z.ai, Grok | 1 credit |
Coding agents cost 5 credits per run because they execute real code, navigate documentation, make tool calls, and recover from errors — a significantly more compute-intensive process than querying an answer engine for a text response.
## Why coding agents cost more
When Lightsage runs a prompt through an answer engine, it submits a query and captures the text response. The entire process takes seconds and uses minimal compute.
Running the same prompt through a coding agent is a fundamentally different operation. The agent reads your documentation, decides which SDK or endpoint to call, writes code, executes it, handles errors, and may retry multiple times. Lightsage captures each step of this process — recommendation rate, tool call success rate, error recovery, and completion. That depth of evaluation takes more time and more compute, which is reflected in the 5-credit cost per run.
## Credits per plan
**250 starter credits** on signup. No credit card required. Use them to explore your visibility score, run prompts on one AI platform, and check your Devtool Arena ranking.
**1,500 credits per month**, included with the \$100/month subscription. Credits reset each billing cycle. Top-ups are available if you need more capacity mid-cycle.
Credits do not expire within your billing period. Any unused credits at the end of the cycle do not roll over to the next month.
## Estimating your monthly usage
Use this formula to estimate how many credits you'll need each month:
```
Monthly credits = platforms × prompts × runs per month
```
Break it down by what you're tracking:
**Prompt Tracking (answer engines and coding agents)**
```
credits = number of platforms × number of prompts × monthly run frequency
```
**API Performance (coding agents and models)**
```
credits = (agents × endpoints × run frequency) + (models × endpoints × run frequency)
```
### Example calculations
Tracking 20 prompts on 2 AI Search platforms, running daily (30×/month):
```
20 prompts × 2 platforms × 30 runs = 1,200 credits/month
```
At 1 credit per run, this is well within the Pro plan's 1,500 monthly credits, leaving 300 credits for API Performance evaluations.
Testing 5 API endpoints with 3 coding agents, running daily (30×/month):
```
5 endpoints × 3 agents × 30 runs × 5 credits = 2,250 credits/month
```
This exceeds the Pro plan's 1,500 monthly credits. To stay within budget, reduce to alternate-day scheduling (15×/month) or narrow to 2 agents:
```
5 endpoints × 2 agents × 30 runs × 5 credits = 1,500 credits/month
```
Combining answer engine prompt tracking with API Performance evaluations:
* 20 prompts × 2 AI search platforms × 30 runs = 1,200 credits
* 5 endpoints × 1 coding agent × 30 runs × 5 credits = 750 credits
**Total: 1,950 credits/month** — slightly over the Pro allocation. Use alternate-day scheduling for lower-priority platforms to reduce usage.
## Tips for getting the most out of your credits
**Start with your highest-priority coding agents.** Claude Code and GitHub Copilot typically have the largest developer audiences. Run them daily and shift lower-priority agents to alternate-day schedules.
**Focus prompts on your top use cases.** If you have 50 possible prompts to track, start with the 10–15 that represent your most competitive categories. You can expand once you've established baselines.
**Use alternate-day scheduling for lower-priority platforms.** Halving run frequency halves your credit spend on that platform without losing meaningful trend data. Most visibility changes are gradual, not overnight.
## Credit top-ups
If you exhaust your monthly credits before the billing cycle resets, you can purchase additional credits directly from your account on the Pro plan. Top-ups are not available on the Free plan.
To discuss custom credit volumes at scale, [book a demo](https://cal.com/lightsage/demo) to explore Enterprise pricing.