EdgeKit AI Access & Answers

Your journalism in Claude and ChatGPT, on your paywall's terms

Readers ask their assistant and get answers from your reporting, decided by rules you write: full text for subscribers, an excerpt and a link for everyone else. A reader who connects without an account registers with you.

The choice today

Readers ask their assistant now. The usual answers cost you the reader.

Block it

The reader still asks. The answer just doesn't come from you, and neither does the reason to subscribe.

Build your own chatbot

A second index, a second login and a second copy of your access rules, drifting from the paywall you already run.

License the archive

The AI company gets the text and the relationship. You get a fee.

AI Access & Answers is the fourth answer: your title inside the reader's own assistant, signed in to your reader account, with your own rules deciding every article.

What it does

Your reporting, inside the reader's assistant

Subscribers get full articles in the assistant they already use. Everyone else gets an excerpt and the article's address on your website, where your paywall makes the offer.

Access

Paywall-grade rules, article by article

Two decisions sit in EdgeKit's decision editor, the same one EdgeKit Paywall uses: who may connect, and what each article returns. Full text, an excerpt, or just the headline and link. Rules read the reader's subscriptions and the article's section, tags, authors, type, language and publish date.

  • An excerpt is the article's standfirst or opening lines, never the passage that matched the question.
  • Access is checked on every search and every read, so a lapsed subscription loses full text on the next question.
  • Publish a rule change and it reaches assistants in 60 seconds, as it does your paywall.
Decision editor · the two AI decisionsPublished

Who may connect Checked on every question

01Signed in · registered or paidAllow

Anyone else is refused.

What each article returns Checked for every article

  1. 01Section is WireHeadline and link
  2. 02Has access to MarketsSection is MarketsFull text
  3. 03Tag is Free to readFull text
  4. ElseOtherwiseExcerpt
Full text
The whole article
Excerpt
Standfirst or opening lines
Headline and link
No text
A common start: any signed-in reader may connect; full text with access to the title, otherwise an excerpt.

Sign-in

Their assistant, your login, one reader record

Readers add your title's address to Claude, ChatGPT or any assistant that takes MCP connectors, then sign in on a page under your title's name with a code sent to their email. An existing reader lands on their own account and subscriptions. A new address registers, and you meet a reader you did not have.

  • Each title's server has its own rules and its own logins. A login for one title is refused by another.
  • The reader is told the assistant sees what they could read on your website themselves.

Your data

The index lives in your database

Your synced archive and its search index sit in your organization's own database, beside your reader records, never in a shared store. Every model call goes through a gateway with article-text logging switched off. When EdgeKit runs in your own cloud account, that gateway runs there too.

  • The assistant receives only what your rules reveal to that reader.
  • Your archive is never pooled with another organization's.

Capabilities

Also in the box

Hide a story entirely

A rule can drop an article from results altogether, not just shorten it.

Any MCP assistant

Assistants register themselves when a reader adds your address, so a new one needs no setup from you.

Abuse limits

Searches and reads are rate-limited per reader, and sign-in codes and app registration per network address.

Search models picked with you

Embedding and ranking models chosen for you. Or rank results with EdgeKit Personalization instead.

Getting live

From archive to assistant in four steps

Your CMS stays the source and your website templates don't change.

White-glove onboarding: we set it up with you, start to finish.

  1. 01

    Connect your archive

    EdgeKit connects to your Arc XP organization and syncs every story. A small integration in your Arc organization tells EdgeKit when a story changes, and an optional nightly sweep catches the changes those events miss.

  2. 02

    We index what's published

    Published stories are split into passages and indexed in your own database.

  3. 03

    Write the two rules

    Write them in the decision editor like any paywall rule. A common start: any signed-in reader may connect and gets full text with access to the title, otherwise an excerpt. Then publish.

  4. 04

    Give readers the address

    Put each title's address on your subscriber pages or in a newsletter. Readers add it to their assistant, sign in once and start asking.

Beside your stack

Adds a surface. Replaces nothing.

Your CMS

Stays the source of record. EdgeKit keeps a synced copy to search and never edits a story.

Your paywall

Same decision editor, subscriptions and meters as EdgeKit Paywall, and one publish takes changes to both live.

Your sign-in

Readers sign in with EdgeKit ID: one reader account on your website and in their assistant.

The reader's assistant

Claude, ChatGPT or any assistant that supports MCP connectors with sign-in. Readers bring their own.

Questions

Asked before every demo

Which AI assistants can readers use?

Claude, ChatGPT and any assistant that supports MCP connectors with sign-in. A reader adds your title's address as a connector, signs in once with an emailed code and allows access.

What does a reader without a subscription get?

Whatever your rules say. A common setup gives the headline, the link and an excerpt, which is the article's standfirst or opening lines. You can tighten that to headline and link, or open chosen stories in full.

Can the assistant get the paywalled passage that answers the question?

No. An excerpt is always the article's standfirst or opening lines, never the passage the search matched. Full text goes only to readers your rules allow.

Will drafts or embargoed stories reach an assistant?

No. Only published stories are indexed or returned. Drafts are never sent to a model, and scheduled stories are never returned.

Do we need EdgeKit Paywall as well?

No. The assistant's two decisions run on their own. Readers sign in with EdgeKit ID, and the rules read the subscriptions and meters EdgeKit holds, which is why it pairs naturally with EdgeKit Paywall and Subscriptions.

Can each of our titles have its own?

Yes. Each title gets its own server, with rules you set per title, and a sign-in to one title's server is refused by another's. Each title needs its own EdgeKit address on its domain, and we set that up with you.

EdgeKit AI Access & Answers

See your archive answer in Claude, under your rules

Book a demo