Paste a city URL. Get a working resident assistant in under five minutes.
Trusted by clerks & communications teams at
Retrieval-augmented generation is the easy part. These are the constraints that make it safe to put on a .gov domain.
Every cited claim traces to a real indexed page.
…residential fences up to 6 feet are permitted as-of-right [1] Zoning Code §4.2.
Citations are validated against retrieved chunks before the answer is shown. Unverifiable claims are stripped.
Below the confidence threshold, it refuses.
I don't have a confident answer for that in the current index. For tax appeals, contact the Treasurer's Office at (703) 435-6800.
No fabricated answers. No hallucinated phone numbers. Residents are pointed to a real human when the index can't help.
Every Q&A logged with its sources for staff review.
Staff dashboard shows every question, the answer given, sources cited, and whether the resident clicked through.
Recrawled on schedule, refreshable on demand.
Pushed a new ordinance this morning? Hit refresh, the assistant catches up in minutes — no support ticket needed.
Built to clear procurement review. No surprises in the data flow, no claims we can't back up.
We only index pages reachable from your public homepage. No intranets, no third-party scraping, no APIs without your explicit configuration. Our crawler respects robots.txt and identifies itself in the user-agent string so your IT team can see exactly what touched your site.
The assistant never requests a resident's name, address, account number, or any personal information. Questions and answers are logged anonymously for staff audit. No IPs or session identifiers are retained beyond short rate-limit windows.
Every connection — resident to widget, widget to API, API to database, API to OpenAI and Anthropic — runs over TLS. Indexed content is scoped per customer with strict access controls; only server-side service credentials read or write, and they rotate on demand.
Your indexed pages are your data. Delete on request and your content is removed from our database. We don't train AI models on your content, and the vendors we use for embeddings and generation operate under commercial API terms that exclude API requests from training.
From the moment you paste a URL to a working assistant on your homepage — under five minutes for most sites.
We crawl public pages, PDFs, ordinances, and meeting minutes. Nothing private, nothing behind a login.
Semantic search and citation-grounded answers, ready in minutes. Every response links back to a source page.
One script tag. Works on any city website, internal portal, or staff intranet. Inherits your typography.
Drop this anywhere on your site. No build step, no npm, no framework required.
<!-- Drop into your site's <head> or before </body> --> <script src="https://cdn.clerkapi.com/widget.js" data-city="herndon-va" data-theme="auto"></script> <!-- Or render inline anywhere on the page --> <div id="clerk-assistant"></div>
ClerkAPI deflects the questions you've answered a thousand times so you can focus on the ones that actually need a human.
Trash pickup, office hours, parking permits — the assistant handles them so your inbox doesn't have to.
Surface the exact application, fee schedule, and submission instructions for any permit type.
Replace the broken site-search nobody trusts with grounded answers that link straight to source pages.
Flat monthly pricing. No per-seat fees, no per-question metering, no procurement maze.
For a single municipality getting started.
For full municipal sites with regular updates.
For counties, multi-city deployments, and state agencies.
Everything we get asked on procurement calls, answered plainly.