RELAY TRANSIT AUTHORITY · FARE
Day pass
For seeing it work
$0forever
- 1 agent on shift
- 50 runs a month
- Full receipts on every run
- Community line
RELAY
--:--:--LOCAL
RELAY · CENTRAL DISPATCH
Your busywork,now departing.
Today an agent runs yourINVOICE TRIAGE
Relay agents pick up the jobs you drop — watch your tools, reason through the mess, and run each one to done. Every run prints a receipt. Approval where it counts, silence everywhere else.
01Departures
Every job on the board. None under the rug.
Your agents' work is a public departures board, not a black box. Watch runs queue, depart, wait for sign-off, and land — in real time.
SCHEDRUNAGENTTASKSTATUS
38 emails drafted, 5 flagged for a human — receipt attached.
112 pairs matched, 3 mismatches escalated with evidence.
Reading 61 new tickets, routing by root cause, not keywords.
12 auto-cleared. 2 over policy limit, waiting on your sign-off.
Departs when the data warehouse finishes its nightly load.
27 accounts scored overnight; drafts staged for review.
2/6 ARRIVED THIS HOURTAP A ROW FOR THE MANIFEST · FULL RECEIPT ON EVERY ARRIVAL
02The console
One window. Every run.
This is the cab, not a mockup of one. Queues on the left, the shift in the middle, and a live tail off whichever run you're watching. No dashboards to build — the board is the product.
RELAY CONSOLE v2.4LIVE · SHIFT 03 · 13:45 LOCAL
TIMERUNTASKSTATUS
13:38RUN-4820REFUND QUEUEDONE
13:41RUN-4821INVOICE TRIAGEDONE
13:42RUN-4822PO ↔ INVOICE MATCHINGRUNNING
13:45RUN-4823TICKET BACKLOGAPPROVAL
—RUN-4824WEEKLY DIGESTQUEUED
QUEUE CLEARS ~17:0099.2% RECEIPTS CLEAN
03The route
Four stops between “someone should” and done.
A Relay run is a relay in the literal sense: your tools hand the job to an agent, the agent hands the risky parts to you, and you hand back a single tap. The baton never gets dropped, because every handoff is on the record.
MEDIAN RUN 4M 12S · CHECKPOINTS ARE YOURS TO SET
STOP 1
Pickup
An agent watches your tools — inbox, CRM, ticket queue, warehouse — and claims the job the moment it appears. No forms, no prompts, no babysitting.
STOP 2
The run
It reasons through the mess: reads the thread, checks the policy, pulls the numbers, drafts the work. Every step is logged as it happens.
STOP 3
Checkpoint
Anything past your thresholds — money, customers, irreversible actions — stops at the gate and waits for a human wave-through. One tap, on your phone.
STOP 4
Arrival
The job lands finished, with a printed receipt: what was read, what was decided, what it cost. Silence unless something needs you.
04Receipts
Trust is a piece of paper.
“The AI did it” is not an answer you can give your auditor, your customer, or your future self. Every Relay run prints a receipt — what it read, what it decided, who approved it, what it cost — kept for seven years, exportable in one click.
- Full step log on every run, kept 7 years
- Named human sign-off on any risky call
- Cost per run, down to the API call
RELAY
RUN RECEIPT · NO. 4822
TUE 09:15 — 09:41 · AGENT: LEDGER
- 09:15:04PICKED UP · REFUND QUEUE (14 CASES)
- 09:15:11READ POLICY §4.2 — REFUND LIMITS
- 09:15:38CHECKED 14 ORDERS AGAINST STRIPE
- 09:16:0212 CASES CLEARED UNDER POLICY
- 09:16:052 CASES OVER LIMIT → CHECKPOINT
- 09:41:19HUMAN SIGN-OFF · M. OKAFOR ✓
- 09:41:4414 REFUNDS ISSUED · CUSTOMERS EMAILED
TOTAL26M 40S · $0.31
HUMAN MINUTES SPENT0.4
KEPT FOR YOUR RECORDS · RELAY.RUN/R/4822
18,402
Jobs completed last week
99.4%
On-time arrivals
4M 12S
Median run, pickup to receipt
0
Silent failures. Ever.
05From the dispatch log
“I stopped checking the board after the first month. That's the review: I trust a robot with our refund queue more than I trusted the old rota.”
“The receipts are the feature. Our auditor asked how we control AI actions and we just… handed over the log. Meeting was ten minutes.”
“Friday, 6pm: forty-one runs arrived, two asked me anything. It's the closest thing to hiring I've done without hiring.”
06Fares
One fare. No meters running.
Flat monthly fare, unlimited runs. If an agent idles all month, it still costs the same — like a good employee, minus the salary.
RELAY TRANSIT AUTHORITY · STANDARD FARE
Commuter
For teams with a queue
$149per month
- 10 agents on shift
- Unlimited runs
- Checkpoints & approval rules
- Slack + email waypoints
- 7-year receipt archive
RELAY TRANSIT AUTHORITY · FARE
Charter
For regulated rails
Customannual
- Unlimited agents
- Self-hosted or private cloud
- SSO, SCIM, audit exports
- A human on our end, on call
07Information
What can an agent actually do?
Anything with an API and a policy: triage tickets, chase invoices, reconcile orders, enrich leads, draft outreach, assemble reports. You define the job and the limits; the agent runs the route. If a tool has no API, agents can work a browser too.
What stops it from doing something stupid?
Checkpoints. You set thresholds — spend, customer impact, anything irreversible — and runs stop at the gate until a named human waves them through. Below the threshold it acts; above it, it waits. The default posture is ask, not act.
Where does it run?
Our cloud by default (SOC 2 Type II, data encrypted in transit and at rest), or self-hosted on the Charter fare. Either way, receipts stay yours and export in one click.
How long does setup take?
First agent on shift in about ten minutes: connect a tool, pick a job from the catalog, set your checkpoint thresholds. The first week it runs with training wheels — everything routes through approval until you loosen it.
What happens when a run fails?
It never fails silently. A failed run parks itself on the board as HELD, pages the on-call human with its receipt so far, and hands back everything it learned. No half-finished actions left in your tools.
NOW BOARDING
The queue you've been meaning to get to is a ten-minute setup away from getting to itself. First agent free, receipts included.
Put an agent on shiftNO CARD · SOC 2 TYPE II · SELF-HOST OPTION