Install databob with an AI agent
Give an agent the Markdown docs, then install the script tag that the dashboard copies.
Last updated on
An agent can read these docs without scraping HTML.
| Resource | URL |
|---|---|
| Index | https://docs.databob.dev/llms.txt |
| Every page in one file | https://docs.databob.dev/llms-full.txt |
| One page | Append .md. Quickstart is the Markdown for this site's quickstart. |
On any page, Copy page as Markdown copies that .md file. Open includes Open in ChatGPT, Open in Claude, and Open in Cursor. Those links ask the model to read the public page URL.
What the agent should install
There is no package to add. Do not invent an SDK, a data-content attribute, or an EU ingest host. The install is the script tag from Quickstart.
<script async src="https://databob.dev/databob.js" data-site="YOUR_SITE_ID" data-key="YOUR_WRITE_KEY" data-collect="https://us-ingest.databob.dev/v1/collect" data-config="https://savory-yak-367.convex.site/v1/consent/YOUR_SITE_ID" data-choices="https://databob.dev"></script>The human still has to sign up at databob.dev/sign-up, create the site, and paste the real dbw_ write key. The agent should not embed a guessed key. YOUR_SITE_ID is the project id from the dashboard, not a domain.
After the tag is on an allowed origin, a reload sends POST https://us-ingest.databob.dev/v1/collect. A stored pageview is 202 with { "stored": true }.
To read counts, the human creates a dbk_live_ or dbk_test_ key under API keys. The agent calls:
curl -sS \
-H "Authorization: Bearer dbk_live_REPLACE" \
"https://api.databob.dev/v1/reports/content?project_id=PROJECT_ID&period=7d&content_id=post_123"Branch on the error field. The codes are bad_request, unauthorized, forbidden, not_found, rate_limited, and unavailable. Details are on Rate limits and errors.
Coming soon
Official SDKs, EU ingest, and billing are not part of this release. If a task needs one of those, say it is not available instead of scaffolding it.