In Gmail
Shared inboxes and your own
Varley works the inbox your team already lives in. It tags and summarises every email, drafts in your voice, answers questions and resolves conversations.
See it working in GmailVarley · AI for customer support
Varley is the AI that does the support work. It reads every conversation, answers from what your business knows, acts in your tools and resolves what it can. It already works in Gmail. Now put it to work in your help desk, your product or your own AI agent.
Where Varley works
Varley is the layer underneath. Put it wherever your customer conversations already are.
Shared inboxes and your own
Varley works the inbox your team already lives in. It tags and summarises every email, drafts in your voice, answers questions and resolves conversations.
See it working in GmailBehind the tool you already run
Put Varley on the queue in the help desk you already use. It works the tickets there, resolves what it can prove and hands the rest to your team.
Ask about your help deskThrough the REST API
Build support into your own app or workflow. Hand Varley a conversation and it resolves it, or hands it back with everything it found.
For developersOver MCP, from Claude or ChatGPT
Give the agent you already use a support colleague. It calls Varley to resolve the conversation, and gets the working back with the result.
For developersAlready resolving real conversations in Gmail, inside Drag and HeyHelp.
Learns from
Acts in
What Varley does
Not a bundle of features. One agent with a method: read first, learn the ground, answer with sources, act only inside the brief, and keep everything on the record.
Let Varley act
Varley working on its own. It reads each conversation you hand it, pulls context from your tools, takes the step and resolves it. When it can’t finish the job, it hands it to your team with everything it found.
What Varley knows
Everything your team knows and the tools it works in, connected once: email, help articles, docs, Stripe, Shopify, Slack and more. Varley grounds every answer in it and shows where each fact came from.
Ask Varley
Ask about your conversations in plain English and get answers from your real data, not a guess. When Varley proposes a change, you see exactly what will happen and approve it first.
Varley on every message
Before anyone opens a conversation, Varley has tagged it by intent and urgency, summarised it if it runs long, and drafted a reply in your team’s voice. Your team reviews and sends.
You decide
The same agent, two levels of trust. Start by asking. Grant autonomy when Varley has earned it.
Ask it
Reading and searching happen instantly. Anything that sends, moves, edits or reaches a customer waits for a person’s approval, with a preview of exactly what will happen.
Let it act
Agent works inside the brief you set. When it can finish the job, it does. When it can’t, it hands the conversation to your team with everything it found.
Most agents treat escalation as failure. Varley treats it as judgement. Knowing what it shouldn’t handle is half of what makes it safe to trust with customers, and you never pay for a hand-off. The rules it keeps are published below.
For developers
The same engine that resolves conversations in Gmail today. Early access opens it to you: connect it over MCP from the agent you already use, or call the REST API from your own product.
The rules it keeps
An agent you can trust with customers is defined by its boundaries. These are Varley’s, in writing.
Every answer is grounded in your knowledge, with the source cited. When the source doesn’t exist, Varley says it doesn’t know and routes to someone who does.
Varley signs its own work. Your customers always know whether they’re talking to the agent or to your team.
What Varley may do is set by you, action by action. The AI can’t widen its own brief, and it escalates the moment a conversation steps outside it.
If a step fails, Varley stops and says so in plain words. Every hand-off is explicit and every action is on the record.
Pricing
You pay for the work Varley does, and only for work you can check. Nothing per seat, and a hand-off to your team is always free.
| Action | Price |
|---|---|
| Resolve a conversationFinished end to end, with nobody on your team involved | $0.50 |
| Ask VarleyA question or task, including the actions it takes | $0.10 |
| Draft a replyIn your team’s voice, with its sources cited | $0.025 |
| Summarise a conversationA long thread, in a few lines | $0.025 |
| Triage a queueWhat needs attention first, across a whole queue | $0.025 |
| Answer a help center searchAn AI answer for a customer, from your articles | $0.005 |
| Tag a messageIntent, urgency and sentiment | $0.001 |
What counts as a resolution is written down in full, dated and versioned. Full pricing →
Varley is an AI that does customer support work. It reads each conversation, answers from what your business knows, takes actions in your tools, and resolves the conversation or hands it to your team with everything it found.
In Gmail today. In early access, in the help desk you already run, in your own product through the REST API, and in your own AI agent through MCP.
One conversation Varley finished on its own, with the customer’s question actually answered, nobody on your team involved, and its working shown. A hand-off isn’t a resolution, and neither is a customer going quiet. The full definition is here.
It hands the conversation to your team and says why, with everything it found attached. Hand-offs cost nothing.
Yes. Autonomy is granted, not assumed. You set the brief, and sensitive actions can require a person’s approval before they happen.
From the places your answers already live: email, help center articles, Google Drive, Notion, Confluence, files and websites. It acts through the tools you connect, such as Stripe, Shopify, Zendesk, Slack and HubSpot.
Teams who want Varley in their help desk, their own product or their own AI agent. Tell us what you want it to handle and we will set you up.
Early access to Varley in your help desk, your own product or your AI agent. Leave your details and we’ll be in touch to set your team up.
✓ Thanks. You’re on the list.
We’ll be in touch at to set your team up.
Prefer email? support@dragapp.com