katman / Integrations
EXPORTS · ANALYTICS · WHAT IS NOT BUILT
Integrations
Most integrations pages are a wall of logos. This one is a list of ways your data leaves katman, a list of the connectors that do not exist yet, and the analytics report you can run today without either. If you came here to check whether we plug into something specific, the second table is the fastest answer.
How your data gets out
| What | What it is | What it is for |
|---|---|---|
| Shareable report link | A signed URL for the run. Anyone with the link reads the full report — every answer, its citations, the model version and the timestamp. No account, no login wall. | Send it to a colleague, a board, or an agency without giving them access to your workspace. |
| The same report as a file. Generated from the stored run, not re-rendered from a screenshot, so the numbers cannot drift from the ones on the page. | Paste into a deck, attach to an email, keep for the record. | |
| Observation log as Markdown | Every dated observation as plain Markdown, ready to paste into Notion, Linear, a repository or a doc. | Keeping the record where your team already works, instead of in one more dashboard. |
| Workspace JSON | The raw workspace: audits, runs, jobs and observations, as one dated JSON file. | Loading into a spreadsheet, a warehouse, or your own scripts. It is your data and it leaves in a format a machine can read. |
| Traffic beacon | One script tag on your own site. It classifies each visit — model referral, AI crawler, search, direct — and writes one row. The key in the snippet is public by construction and only works for the hostnames you allowlist. | Seeing AI-referred visits in your own numbers rather than watching them disappear into “direct”. |
Everything above works on the free audit as well as a paid run, except the traffic beacon, which needs a key issued against your hostnames.
What we do not have
A vendor's integrations page is usually where the overclaiming happens, so this table is here first and the caveats are not in a footnote.
| Not built | Why we are saying so |
|---|---|
| A public read API | Not built. The only public endpoint is the write-only traffic collector. Workspace JSON is the export path today. |
| Native GA4, Looker or BigQuery connectors | Not built, and we would rather say so than call a CSV a connector. The measurement below is a report you run yourself in GA4 in about a minute. |
| Zapier, Make or webhooks | Not built. Nothing pushes out of katman on an event yet. |
| Slack or email alerts on movement | Not built. It is on the list, and it is honest to say it is a list and not a roadmap with dates. |
| A CMS plugin that publishes for you | Deliberately absent. We never publish or post anything as you — the work list is yours to ship, and that is a product decision rather than a missing feature. |
Read the AI channel in the analytics you already run
You do not need a connector for this, and you do not need us. It takes about a minute and it is the single measurement most companies are missing — in our own first-party study, the largest channel on the site was hiding inside direct for a month.
01 Find the model referrals you already have
GA4 › Reports › Acquisition › Traffic acquisition, then search the Session source / medium dimension for
chatgpt.com,perplexity.ai,claude.aiandgemini.google.com. Assistants that pass a referrer show up as referral rows; the ones that do not land in Direct.02 Read the tagged sessions
Some assistants append
utm_source=chatgpt.com. In a server log that is one line:grep -E 'utm_source=chatgpt\.com' access.log | awk '{print $1}' | sort -u | wc -l. In GA4 it is the same source dimension filtered to that value.03 Join it to revenue, not to sessions
Sessions are the easy half and the one people quote. Take the orders in your payment provider for the same window and trace each back to its first tagged session. This is the step that turned a share-of-traffic number into a revenue number in our own study, and it is the step that most attribution skips.
04 Accept what you cannot see
AI crawlers do not run JavaScript, so no browser-based analytics — ours included — will ever show you GPTBot or ClaudeBot. That half comes from server logs or your CDN, and any tool claiming otherwise is guessing.
Where katman sits next to what you have
katman is not a replacement for your analytics, your rank tracker or your CMS, and it does not want their jobs. Analytics tells you what visitors did after they arrived. A rank tracker tells you about a results page. katman answers a different question: when a buyer asks a model before they have heard of you, does it name you, and who does it name instead. The output is a dated record and a list of pages to publish, which then goes into whatever system your team already ships from.
Need something that is in the second table? Say which one on the contact page. We would rather hear that an export format is blocking you than guess at a roadmap.