> ## 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.

# Track prompts across answer engines and coding agents

> Learn how Lightsage runs prompts across answer engines and coding agents, what each run measures, and how to read the results.

Prompt tracking shows how AI platforms describe, recommend, and rank your API over time. Lightsage runs your saved prompts across answer engines and coding agents on a schedule, then turns the responses into visibility metrics, competitor comparisons, cited sources, and actionables.

Use prompt tracking when you want to answer questions such as:

* Which AI platforms mention your API for important developer use cases?
* Which competitors appear when your API does not?
* Do coding agents recommend you when a developer asks how to build something?
* Which docs pages or content gaps should you fix next?

## What Lightsage tracks

Each tracked prompt belongs to a topic and runs against one or more AI platforms. A topic usually maps to a product category, use case, framework, or competitive area.

For each run, Lightsage records:

| Result | What it tells you |
| - | - |
| **Mention presence** | Whether your company or product appears in the response. |
| **Position** | Where your brand appears when the response ranks or lists options. |
| **Share of Voice** | How your mentions compare with competitor mentions for the same prompt set. |
| **Citations and sources** | Which pages the platform cites or appears to rely on when sources are available. |
| **Recommendation quality** | Whether the response recommends you for the user's stated intent. |
| **Actionables** | Suggested content or documentation fixes based on the result. |

Prompt tracking is separate from API Performance. Prompt tracking measures whether AI platforms recommend and describe you. API Performance measures whether coding agents can successfully use your API, SDK, CLI, or MCP server in real tasks.

## Platform types

Lightsage separates prompt runs into answer engines and coding agents because they influence different parts of developer adoption.

> **Answer engines**
>
> Answer engines respond in prose. Use them to measure awareness, category positioning, comparison coverage, and source citations.
>
> Examples include ChatGPT, Perplexity, Claude, Gemini Search, Microsoft Copilot, Google AI Overviews, and Google AI Mode.

> **Coding agents**
>
> Coding agents work inside developer workflows. Use them to measure whether agents recommend your API when a developer is ready to implement.
>
> Examples include Claude Code, OpenAI Codex, Cursor, GitHub Copilot, Gemini CLI, OpenCode, OpenClaw, Hermes, Pi, and Kilo.

Answer engine runs cost 1 credit per platform run. Coding agent runs cost 5 credits per platform run because they use agent execution paths rather than a single answer-engine request.

## Choose prompts to track

Track a focused set of prompts that reflect the decisions developers make before and during implementation.

| Prompt type | Use it to measure | Example |
| - | - | - |
| **Brand** | Whether AI systems know who you are. | `What is Acme API?` |
| **Category** | Whether you appear for your market or use case. | `Best payment API for a SaaS app` |
| **Comparison** | Whether you win head-to-head evaluations. | `Acme API vs Stripe for marketplace payments` |
| **Implementation** | Whether coding agents recommend you during build tasks. | `Add Acme payments to my Next.js app` |

Start with 3-5 brand prompts, 5-10 category or comparison prompts, and 3-5 implementation prompts for your most important frameworks or use cases.

> [!TIP]
>
> Write implementation prompts the way a developer would ask a coding agent. Include the framework, runtime, or integration target when it matters.

## Set up tracking

**1. Create topics**

Group prompts by product area, category, framework, or competitive set. Topics make reporting and Share of Voice comparisons easier to read.

**2. Add prompts**

Add prompt text for the questions you want Lightsage to run. Keep each prompt tied to one clear intent.

**3. Select platforms**

Choose answer engines for awareness and source coverage. Choose coding agents for implementation-oriented prompts where adoption is the goal.

**4. Choose a schedule**

Run high-priority prompts daily when you are actively changing docs or positioning. Use alternate-day runs for stable prompts where cost efficiency matters more than immediate feedback.

**5. Review results**

Use Visibility Analytics to compare Visibility Score, Share of Voice, average position, platform breakdowns, citations, and prompt-level trends.

You can manage prompts in the dashboard or through the CLI.

```bash theme={null}
lightsage prompts config retrieve
lightsage prompts list --output-format table
```

Create prompts through the CLI when you want a repeatable setup:

```bash theme={null}
lightsage prompts create \
  --prompts '[{"text":"Best payment API for a Next.js marketplace","topic_id":"topic_123","platform_ids":["chatgpt"]}]'
```

## Read the results

Start with the trend, then drill into the prompt and platform that changed.

| View | Use it for |
| - | - |
| **Visibility Score** | Track the overall direction of your AI visibility. |
| **Share of Voice** | See whether competitors are gaining or losing ground. |
| **Prompt rows** | Find the exact prompts where you win, lose, or disappear. |
| **Platform breakdown** | Separate answer-engine awareness problems from coding-agent adoption problems. |
| **Citations** | Identify which docs, comparison pages, or third-party sources shape the answer. |
| **Actionables** | Pull prioritized fixes into your docs or product workflow. |

Low scores on answer engines often point to category, comparison, or source-authority gaps. Low scores on coding agents often point to implementation documentation, missing examples, weak `llms.txt` coverage, or unclear SDK and API guidance.

## Turn tracking into action

After you identify a losing prompt, use the result to decide the next fix:

* If competitors are mentioned and you are not, add or improve category and comparison content.
* If agents cite the wrong page, update the canonical docs path and make the answer easier to extract.
* If implementation prompts fail, improve quickstarts, SDK examples, endpoint docs, and `llms.txt` coverage.
* If a prompt is important but noisy, rewrite it around a narrower framework, use case, or buyer decision.

For agent-ready follow-up work, use [Pull actionables into your agent](https://lightsage.com/docs/workflows/pull-actionables-into-your-agent.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.