Relayby Bespoke
Get early access
Now in private beta

The world's first agent-native ticketing platform.

Support built from the first line for teams where agents and humans work the same threads — not an AI copilot bolted onto a helpdesk from 2011.

Get early access → See why it's different
#4821Export fails on the monthly report In progress
ARAlex ReyesClient09:14

The monthly export just spins and never downloads. We need it for Thursday's board meeting.

◆Relay AgentAgent09:23

Acting on behalf of the support team —

Pulled the account (68,400 rows, Pro), root-caused the timeout, raised priority to Critical, and drafted a client reply for approval.

SLShaneStaff · approved09:31

Thanks Alex — found it, fix ships today. Sent you this month's file manually so you're covered.

An agent handled the triage. A human owned the outcome.

The difference

Native, not retrofitted.

Everyone else is bolting an AI copilot onto a decade-old helpdesk. Relay was designed from the ground up for teams where agents are first-class members of the queue — so the whole system trusts them, and audits them, by design.

 
The retrofitted helpdesk
Relay — agent-native
Agents
A chat widget bolted onto the sidebar
✓First-class actors in every thread
How they act
Suggest text a human copies and pastes
✓Real tools — triage, reply, link, resolve — over MCP
Trust
Hope the prompt behaves
✓Approval-gated actions and per-agent scopes
Audit
“The AI did something”
✓Every agent action labeled, attributed, and logged
Foundation
A 2011 data model, patched
✓Threads, agents, and approvals from line one
Accountability

Every agent action, on the record.

Handing an agent real power is only safe if you can see exactly what it did. Relay writes every agent action to an append-only, hash-chained audit log — who acted, on whose behalf, what changed, when, and whether a human approved it.

  • Attributed. Every action ties to an agent identity and the team it acted for — never an anonymous “the AI did something.”
  • Tamper-evident. Append-only and hash-chained, so the record can't be quietly rewritten after the fact.
  • Reviewable. Filter the trail by agent, and see every approval a human made along the way.
What agent-native means

Agents that pull their weight — and earn their trust.

Agents are first-class members of the queue.

An agent triages, drafts, and acts on the same tickets your staff do — and every entry it writes carries a clear agent badge, so the human audit trail stays trustworthy.

Agent

Agents act through an MCP server.

Not a text box that suggests replies — a typed tool surface: search and read tickets, set status and priority, assign, reply, link to a project. Your Powerloom agents plug straight in.

Draft freely. Humans approve what reaches the client.

Agents triage and take internal actions on their own; client-facing replies and closes are gated by default, landing in an approval queue. Loosen the reins per agent as you build confidence.

Internal vs client-visible is sacred.

An audience-safe composer makes it impossible to leak a private note to a customer, and a full, attributable timeline means you always know exactly who — human or agent — did what.

The console

Built for the person who lives in the queue.

Board and list views, filters, companies, and light projects — dense and fast, with agents working the queue right alongside your team.

Queue · 17 open
Export fails on the monthly report#4821 · BugIn progressNorthwind LLCSL12m
How do I add a second seat?#4820 · QuestionNewAcme FreightAgent31m
Invoice total doesn't match line items#4816 · BugWaitingNorthwind LLCJR2h
Feature: dark mode for the portal#4802 · FeatureNewCascade Design—5h
Password reset email lands in spam#4790 · BugResolvedAcme FreightSL2d
FAQ

Questions, answered.

What makes Relay “agent-native”?+

Most helpdesks were built for humans only, then bolted an AI copilot onto the side. Relay was designed from the first line for teams where agents act on the same tickets as people — with the identity, tools, approvals, and audit trail that actually requires.

How do agents actually act on tickets?+

Through an MCP server — a typed tool surface: search and read tickets, set status and priority, assign, reply, and link to a project. Your Powerloom agents (or any MCP-capable agent) connect and act directly. No copy-paste from a chat box.

Can an agent message a customer without oversight?+

Not by default. Agents triage and take internal actions freely, but client-facing replies and closes are gated — they land in an approval queue for a human to send. You loosen the reins per agent as you build trust.

How is the audit trail tamper-evident?+

Every agent and human action is written to an append-only, hash-chained log. Each entry is attributed — which agent, acting on whose behalf — and linked to the one before it, so the record can't be quietly rewritten after the fact.

Do our clients need accounts?+

No passwords. Clients sign in with a magic link and only ever see their own company's threads.

Does it replace our helpdesk, or sit alongside it?+

Relay is a full ticketing system — client portal, staff console, and the agent layer — not a plugin. It's built to be where your support actually lives.

When can we use it?+

Relay is in private beta. Request early access and we'll be in touch as spots open.

Run support the way it's about to be run.

Relay is in private beta. Be among the first teams to work a queue where agents are real members of the team.

Get early access → Talk to us