Docs
What Saylent is, what one run gives you, and the six sections these docs are organized into.
Saylent asks real buyer questions to real AI assistants - ChatGPT, Claude, Gemini and Perplexity, through their official APIs - records every answer verbatim, and tells you whether your brand was recommended, mentioned or absent, which rival was named instead, which pages the assistants cited, and what to fix. It runs on your own provider keys, on your own machine or your own infrastructure. Nothing is sent through us and nothing is stored by us.
To see a full report before you set up anything, run npx saylent demo: it
renders the bundled sample run offline, with no key and no cost.
Which tool?
promptbeacon is for developers and open-source maintainers: does AI recommend your library or dev product — a number in CI and a README badge. Saylent is for brand and website teams: an audit of what AI assistants say about your site, with the answers and cited pages as receipts, plus fixes for your site.
These docs are one story in six parts, in the order people actually need them.
Run it once
One command, your own keys, a report you can open in five minutes. No account, no database.
Keep score as a team
The same engine with a memory: brands, history, a fix tracker, share links, and as many user accounts as you like on one deployment you own.
Operate it for others
The console for whoever owns the keys and the bill: spend cap, kill switch, models per role, users, runs, takedowns.
Automate
A $0 CI gate on every push, a Claude Code plugin and an MCP server so an agent can run it, and the pipeline as a library in your own code.
Understand
What a run does at each of its nine stages, what it costs, what leaves your machine, and what it refuses to claim.
Project
The $0 dev setup, the extension points, the changelog, and where to report a security issue.
| What you need | Where it lives | What it costs |
|---|---|---|
| To see what a report looks like first | The CLI: npx saylent demo | $0, no keys, offline |
| One report on one domain, today | The CLI: npx saylent audit example.com | about $0.80 to $1.20 with four engines |
| To know which AI crawlers your site lets in | The site check: npx saylent gate-check example.com | $0, no keys |
| To know which AI crawlers actually visit | AI crawlers: npx saylent crawlers access.log | $0, no keys, the log stays on your machine |
| To ask what your buyers really search | Import questions: npx saylent questions import Queries.csv | $0, no keys |
| To be told when the answers move | Alerts: saylent verify --notify slack, or per brand in the app | $0 beyond the run itself |
| Your data in a spreadsheet, a dashboard or a script | Export and API: npx saylent export run.json, the app's Export menu, or the read-only API | $0 |
| The drafted fixes on a branch of your site repo | saylent fixes apply | $0; it commits locally and never pushes |
| History, a fix tracker and more than one account | Deploy the app | the same per run, on your own keys |
| A gate that fails a push when a bot is blocked | The GitHub Action | $0, no keys |
| Claude Code to do it for you | The Claude Code plugin | $0 for the site check; the same per run for an audit, after you agree to the estimate |
| Your agent or your own program to run it | The MCP server or the library | the same per run, on your own keys |
Before you spend anything
- What it costs - real numbers per profile, from recorded runs, and the caps that stop a run before it starts.
- What gets sent to providers - exactly what leaves your machine, to whom, and what never does. No telemetry, anywhere.
- Methodology - how presence is decided, why a judge from the other provider family scores the answer, and the limits we state out loud.
- Compared with hosted tools - sourced prices, dated, and the five things the hosted tools still do that Saylent does not.
- Walk the live demo - a read-only copy of the app on the Kestrel Uptime sample: every screen, nothing saves, no account, $0.