Conversations
One live inbox across every workspace you can see.
Routing groupsInternal notesCSAT
Features
Everything in your dashboard, explained screen by screen.
This page is organized the way the product is. Every section below maps to a real screen your team will use — no abstract feature language.
One async scriptYour AI provider keyExportable conversations
docs.runwick.com/webhooksRunwick · onlinePartial refunds fire webhooks?Yes, both.[1] Webhooks → Events
parkvill.com/account/ordersParkvill · onlineChange address on #4821?Already packing. Getting a person.handed offPriya · ParkvillHolding #4821. Send the new address here.
Analytics
Six numbers that tell you if support is working.
Every metric ships with a sparkline and a period-over-period delta across 7, 30, or 90 days — including tokens per conversation, so the AI bill you pay your provider stays visible.
Analytics — acme.com
Conversations+18%
1,284
AI-handled+6 pts
72%
Handoff rate−3 pts
11%
Resolved+21%
1,102
Average CSAT+0.2
4.6
Tokens / conversation−9%
3.1k
When conversations arrive (24h)
Sparkline + delta on every metric7 / 30 / 90-day periodsToken spend stays visible — you pay your provider, not a markup
Knowledge
Teach it from the site you already maintain.
Point a scan at your site, approve what it recommends, and watch each source move from discovered to indexed. Nothing enters the AI's knowledge without you saying so.
Learn from your website
docs.acme.com
Scope: Recommended
Scan site
Recommended knowledge
/guides/getting-startedApprove
/policies/refundsApprove
Sources
+ Add URL or Markdownshipping-policy.mdReadyindexed 2h ago
docs.acme.com/returnsReadyindexed 2h ago
docs.acme.com/api/authIndexing38 chunks
pricing.mdDiscoveredawaiting approval
legacy-faq.htmlFailedReindex
Scan scope: recommended, one section, or the whole sitePer-source status with error text and one-click reindexYou approve every source before it answers
The rest of the map
Eight more screens, each doing one job well.
Canned replies
A snippet library for the answers you keep retyping.
Insert in composerUsage metrics
AI settings
Your provider, your key, your model — encrypted at rest.
5 providersReadiness probe
Skills
Brand voice as settings, not prompt engineering.
Tone & lengthRequired guardrails
Widget
Theme it, place it, schedule it — with a live preview.
Paste this before </body> on every allowed site.
Visibility rulesOffice hours
Site keys
Exact-origin allowlists and signed visitor identity.
Origin allowlistVerified visitors
Team
Owners and agents. Public display names for visitors.
InvitationsTwo roles only
API & webhooks
Scoped tokens, signed webhooks, delivery replay.
REST + GraphQLEvent replay
One line on the site. The whole support loop behind it.
Start with one workspace for one launched site. Add another when you need it.
Workspace-scoped dataExportable conversationsProvider key stays encrypted