For customer service teams

One policy. The same answer, every channel.

Ask an ungoverned bot the same policy question twice and you can get two different refund windows. Your help bot, your agents, and your help center should never disagree. Keep the approved policy in one place and every channel answers from it.

One approved policy behind every channel
Customer-safe and internal kept separate
Every answer traceable to its approver
Free tier to start

Your AI assistant

Sampleanswering from approved knowledge

Asked

Can a customer change their shipping address after an order ships?

The approved answer

No. Once an order leaves the warehouse, the address on file cannot be changed. Offer a reship at the customer's cost or a redirect through the carrier if the carrier supports it.

Grounded in knowledge your team approved

source: policies/shipping-changesapproved by Support leadership

The policy, not a paraphrase

Refunds, entitlements, and escalation rules become approved pages, so the bot quotes the policy your team wrote instead of guessing at the edge case that would have generated your angriest ticket.

Two audiences, on purpose

Your agent-assist copilot and your internal team can read the full playbook; the customer-facing bot reads only customer-safe answers. Scoped keys keep the line between them.

Changes land everywhere at once

When the policy changes and the new version is approved, every connected channel answers with it from the next question on.

Provable when questioned

When a customer says 'your bot told me', you can trace the answer to the exact approved version and who signed it off.

Questions

Does this replace our helpdesk?: No. Tickets keep flowing through your helpdesk. This is the approved knowledge behind whichever bot or assistant answers on top of it.
How is this different from our help center?: A help center is written for customers to browse. This is the reviewed, versioned source your AI answers from, including the internal guidance a public page cannot hold.
Can the bot mention internal-only rules?: No. Customer-facing connections read only what they are cleared to share. Internal escalation thresholds stay with your agents.
What does it cost to try?: Free is $0 and covers 200 pages and 5 members, enough to prove this on your top policies. Team is $99 a month once you are running this across real channels and need more room. Enterprise is custom priced for larger organizations.
Do we have to write everything again from scratch?: No. Import your existing help center or policy docs in bulk, then send only the exceptions and edge cases through review. Most teams start from what they already have, not a blank page.
Will this pass our security review before the bot goes live?: Approval filtering, workspace isolation, and scoped tokens are enforced by the retrieval service itself, not by the bot's prompt. Single sign-on (OIDC) is available on the Enterprise plan; SAML and SCIM are not built yet. If one of those is a hard requirement for your review, we will tell you that directly rather than promise a date.

Rolling this out at scale, across channels?

Tell us how many channels you're connecting and the volume you're running, and we will help you plan the rollout.

Stop shipping guesses to customers

Start free, write down the policies your team repeats most, and connect your bot.

Connect in a minute

Create a token from Settings → API Tokens, then point your MCP client at the hosted endpoint.

.mcp.json

{
  "mcpServers": {
    "curated-data": {
      "type": "http",
      "url": "https://curateddata.megacorp.company/api/curated-data/mcp",
      "headers": { "Authorization": "Bearer ckd_live_••••••••" }
    }
  }
}