For the complete documentation index, see llms.txt. Every page on this site is also served as Markdown: append `.md` to any URL, or send `Accept: text/markdown`.
Attensira Logo
Attensira
For SaaS & Tech Startups

For founders whose docs
already live in a repo.

Attensira reads what ChatGPT, Perplexity, Google AI, Claude and Gemini answer for your category, finds the pages that would change the answer, and opens a pull request against your repo. You review and merge.

Get StartedBook A Demo
•Cancel anytime•Nothing merges without your approval
The startup problem

Three things a dashboard leaves you holding

What early-stage teams run into once buyers start asking models instead of searching.

01

You don't know whether models name you at all

A buyer asks ChatGPT for the best tool in your category. You have no record of what it answered, how often, or which domains it linked. Guessing is the default state, because nobody stores the answers.

02

The same prompt gives a different answer every time

Models are not deterministic. One screenshot proves nothing. The only honest reading is a rate over repeated samples, with the sample size attached and small movements refused rather than reported.

03

Knowing which page is weak doesn't fix the page

A dashboard hands you a finding and leaves the edit to you — the part that takes the afternoon. The page lives in a repo, so the fix should arrive there, as a diff you can read.

5
AI answer surfaces read on a schedule
1/day
Max runs per prompt, model and country
95%
Significance bar before a change is reported
PR
How every fix reaches your repo
Built for founders

What a small team actually gets

Measurement you can check, and edits that arrive as reviewable diffs.

See the answer, not a score

Attensira stores the verbatim text of every answer it collects, so you can read the paragraph that named a competitor instead of yours. There is no blended visibility score to argue with — just mention rate, citation rate, and the sample size behind each.

See how tracking works

Track the buyer questions, grouped by topic

You write the prompts your ICP actually asks — 'best AI coding agent for startups,' 'cheapest CRM for a five-person sales team' — and group them into topics. Each prompt runs against each tracked model, at most once per prompt, model and country per day.

See how tracking works

Score competitors over the answers you already collected

Add a competitor by domain and Attensira scores their mention and citation rates over the same answers, on the same denominator — no extra runs, no prompt slots spent. Rates are independent, so they can sum past 100%: one answer can name four brands.

See how competitor tracking works

Made for founders, not procurement

You install the GitHub App into your own org yourself, scoped to the repositories you pick, and you can cancel anytime. There is no free plan and no trial — you buy, then the agent starts.

See plans

Where a Git-based team goes next

The pull-request path, and the crawler decisions that come before it.

GitHub: the agent opens the pull requestCrawler logs, and what actually fetched your pagesPrompt tracking across the tracked answer surfacesEvery documented AI crawler, with its robots.txt tokenTraining versus search: the two kinds of AI crawlerAI visibility for information and media

Point it at your repo.
Read the first pull request.

Add the buyer questions you want to win, install the GitHub App on the repositories your site and docs live in, and see what the models answer today.

Get StartedBook A Demo
•Cancel anytime•Every PR is yours to reject
FAQ

Frequently Asked Questions

Want to see a real pull request first?
Book a walkthrough and we will run your category live.
Book A Demo

It runs the buyer questions you care about — 'best CRM for a five-person sales team' — against ChatGPT, Perplexity, Google AI, Claude and Gemini on a schedule, stores the verbatim answers, and reports two numbers per question: how often the answer named you, and how often it linked your domain. Then it writes the page edits that would change those numbers and opens a pull request against your repo. You review and merge. Nothing publishes without you.

We do not publish a timeline, because we have no customer results to derive one from. What you get instead is a measurement you can trust: every rate is reported with its sample size, and every change is put through a two-proportion z-test at 95% before we call it a change. If a movement does not clear that bar, Attensira says so rather than drawing you a line that goes up.

The work arrives as pull requests, so reviewing it is the same as reviewing a teammate's branch. Each PR is on a fresh branch, carries the `attensira` label, and its body names the page it edits, the finding it answers, and why. The agent never merges, never pushes to your default branch, and never touches CI config.

Developer tools are the best fit, because your docs are already in a Git repo. Attensira installs as a GitHub App into your own org, scoped to the repositories you pick, with three permissions: Contents read/write, Pull requests read/write, Metadata read. A docs fix lands in the repo the page lives in.

There is no rank to climb. A model either names you in its answer or it does not, so the unit of measurement is a rate over sampled answers, not a position. Attensira also buckets every domain a model cites as your own, a competitor's, an AI platform, a reference site, or a third party — the third-party bucket is the one where a citation is genuinely winnable.

Starter gives you a fixed number of concurrently tracked prompts and one sample per prompt per model per reading. Growth raises sampling to three, which is what makes small differences measurable, and adds Gemini. There is no free plan and no trial — you buy before the agent does anything, and you can cancel anytime. The plan numbers are on the pricing page.