uplink 00:00:00

Inside your network. Callable from anywhere.

The systems that matter most are often the ones you cannot reach: an internal portal, a supplier extranet, a legacy admin screen on the far side of a VPN. Uplink gives them a front door without opening one in your firewall.

Outbound-only · no inbound port · free on your own hardware
how 00:00:10
01
The mechanism

The agent dials out. Nothing dials in.

You run a Writ agent on a machine that can already reach the system — your laptop, a box in the office, a container in the same VPC. It opens one outbound HTTPS connection to Writ and keeps it. That is the whole network change.

  1. The agent connects outward to Writ and stays connected.
  2. A call to your published endpoint is handed down that existing connection.
  3. The workflow runs on your machine, against a system only it can see.

The page never leaves your network. The agent returns the result of the workflow — the rows you extracted, the confirmation you captured — and nothing else.

why 00:00:20
02
What this replaces

The three usual answers, and what they cost.

Every team that needs this has been offered the same three options. Uplink is the fourth.

Open a port

A firewall exception, an inbound rule, a public hostname, and a new thing on your attack surface — for one integration. Security review measured in weeks.

Build the API

Ask the vendor for one, or write a service against a system nobody wants to touch. Both are real projects with an owner and a budget.

Do it by hand

Someone logs in every morning, copies rows into a sheet, and the process stops when they are on holiday.

limits 00:00:30
03
What each plan opens

How many doors, and how hard you can knock.

A door is a published REST endpoint or MCP tool. The daily call ceiling is fair-use on the relay that carries traffic to your agent.

PlanPublished doorsRelay calls / day
Open Source 2 2,000
Starter 5 10,000
Pro 15 25,000
Growth 40 50,000
Scale 100 100,000
Enterprise Unlimited 250,000

Runs dispatched to your own machine carry no compute charge on any plan, including the free tier. You are paying for the front door, never for the work behind it.

venue 00:00:40
04
Where it runs

Two machines, one front door.

The same published endpoint can be served by our fleet or by your agent. Callers see one URL; you choose which side of the wall answers it.

your network nothing opened caller your code · your AI front door api.usewrit.app Writ agent on your laptop

What this required from your firewall: no inbound rule · no public hostname · no port opened

Nothing dials in. The agent holds one outbound connection, and the call rides back down it. There is no inbound rule to write and no port to open.

Your own machine — free, no account, runs while it’s awake; only the result leaves.

The agent dials out. Nothing dials in — a system behind your VPN becomes callable without opening a single inbound port. Local callers can stay entirely local; add Uplink only when a remote caller needs to reach a workflow behind your firewall.

Questions people actually ask 00:00:50
05

Questions people actually ask

Do I have to open a firewall port?
No. The agent makes an outbound HTTPS connection to Writ and keeps it open; calls travel down that existing connection. There is no inbound rule, no port forward, and no public hostname for the machine.
What actually leaves my network?
The output of the workflow you recorded — the fields you chose to extract, or a confirmation that an action completed. The pages themselves, the session, and the credentials stay on the machine running the agent.
What does it cost to run?
Nothing for compute. A run on your own machine is unmetered on every plan, including the free tier, and that holds whether you started it yourself or it was dispatched from the cloud. Your plan governs how many doors you can publish and the fair-use ceiling on relay calls.
Can I self-host the whole thing?
Yes. The open-source build runs the agent and the engine on your own hardware. The billing code is not compiled into it, so it cannot charge you.
What happens if the agent is offline?
The call fails with a clear error rather than hanging — there is no machine on the other side to serve it. Link more than one agent if a workflow needs to answer while any single machine is down.

Give your internal system a front door.

Free on your own hardware, forever. No inbound port, no public IP.