Drag AI

Agents

Drag Agent reads every email, classifies by intent, pulls context from your tools, takes real action, and resolves - or escalates when it should.

Email in. Resolution out.

Four steps, under 30 seconds. Drag Agent classifies intent, retrieves context from your tools, takes real action, and resolves - or escalates to your team with everything it learned attached.

Start free trial Book a demo

Included in Plus ($18/mo) and Pro ($24/mo). No per-seat AI add-on.

agent trace

09:41:02CLASSIFYintent=billing, entity=invoice #4821

09:41:03RETRIEVEstripe.invoices.get(4821) → $249.00 paid

09:41:04RETRIEVEkb.search('refund policy') → 30-day window

09:41:05ACTstripe.refunds.create(amount=249.00) → rf_8xK2

09:41:06DRAFTconfidence=0.96, sources=2, tokens=142

09:41:07SENDthread_id=t_9f3a → resolved

23s avg resolution0 unsourced answers

How it works

It resolves. Or it hands off.

Most email AI stops at a suggested reply. Drag Agent finishes the job: it reads the incoming email, works out what the customer actually needs, pulls the records that answer it, executes the fix in the tools you connect, and sends the response - with every step logged and every claim sourced. And when a conversation is outside its depth, it does the other professional thing: it escalates to your team, context attached, before anything ships.

Two conversations, two calls

Refund status questionconfidence 0.96

Policy found, payment record matched → resolved & sent

Custom contract amendmentconfidence 0.41

Below threshold → escalated to Sarah with context attached

The threshold is yours to set. Nothing ships below it.

Judgment built in

Knows when to act. Knows when to ask.

Every resolution carries a confidence score, and nothing ships below the threshold you set. High confidence resolves autonomously; anything uncertain escalates to a teammate with the full context the agent gathered - so your team starts from the middle, not from scratch.

Connected

Connected to every tool your team uses.

The agent reads from your knowledge base, docs, and PDFs; acts in the systems that hold the answer - Stripe, HubSpot, Zendesk, Jira and more; and listens on the channels your customers actually use. Intelligence, Drag's knowledge layer, grounds all of it.

Reads from

Knowledge baseNotionPDFs & URLs

Acts in

StripeHubSpotZendesk

Listens on

GmailSlackWhatsApp

Grounded by Intelligence, your connected knowledge layer

Resolution tracesearchable · timestamped

09:41:02CLASSIFYintent=billing, invoice #4821

09:41:04RETRIEVEkb.search('refund policy') → 30-day window

09:41:05ACTstripe.refunds.create → rf_8xK2

09:41:06DRAFTconfidence=0.96, sources=2

09:41:07SENDthread t_9f3a → resolved

Trust

Every action logged. Every source cited.

The agent keeps a full audit trail: every classification, retrieval, API call, and send, timestamped and searchable. Every response links back to the article, record, or doc it drew from. SOC 2 and CASA certified, encrypted at rest and in transit, no training on your data.