How installing works on the Writ Marketplace
Installing on Writ isn't a download or an integration project. You install a frozen, data-less recipe into your own workspace; you attach the login, secrets, and inputs it needs; and you run it on your machine or our cloud. Here's each step, and exactly what a run costs.
Discover
Find listings by full-text search and filters - success rate, latency, reputation - or browse by category, curated collection, capability set, or creator storefront.
Install a recipe, not the creator's data
Install drops a frozen copy of the recipe into your workspace: the steps plus a manifest listing the inputs, secrets, and personas it needs. The creator's own credentials, sessions, and secrets were stripped at publish - what you install can't contain them. Your install stays on the version you tested until you choose to sync an update.
Attach your own data
Map the manifest to your data: pick your own persona or login, reference your own vault secret, and fill in your input values. A creator can attach a regex rule to any input or secret slot - your value must fully match it or the run won't start, which catches a malformed value before the run begins.
What you attach
- Your persona / login: Including built-in 2FA - TOTP and mailbox-OTP - for login-gated sites.
- Your vault secret: Referenced from your vault, never typed into a step.
- Your input values: The values the workflow runs against.
Per-input regex rules
A creator can require a value to fully match a pattern before the run starts.
rule: ^https?://(www\.)?amazon\.
A failing value stops the run before it starts.
Run on your account, your way
Run it on your own local agent - which can reach intranet, localhost, and VPN-only systems - or on our managed cloud fleet with nothing to install, within the creator's run policy. Either way it's callable as a REST endpoint and an MCP tool, authenticated with your own wt_ keys.
curl -X POST https://api.usewrit.app/v1/amazon-price-watcher/run \
-H "Authorization: Bearer wt_xxxxxxxxxxxxxxxxxxxx" \
-H "Content-Type: application/json" \
-d '{
"product_url": "https://www.amazon.com/dp/B0XXXXXXXX",
"target_price": "59.99"
}' {
"mcpServers": {
"writ": {
"url": "https://api.usewrit.app/mcp/amazon-price-watcher",
"headers": {
"Authorization": "Bearer wt_xxxxxxxxxxxxxxxxxxxx"
}
}
}
} What a run actually costs
Every listing is free to install and free to run. The only thing a run costs is Writ platform compute, and only when the run happens on our managed cloud.
| Layer | On your own machine | On the managed cloud |
|---|---|---|
| The listing itself free to install, free to run | Free | Free |
| Writ platform compute the only thing a run can cost | Free | Metered by running time |
Run it on your own machine and it costs nothing. A local or BYO run is free and unmetered, on every plan. On the managed cloud, compute is metered by running time at about $0.12/browser-hour, drawn from your plan’s included usage.
No surprise compute bill - pre-flight wallet check + credit-hold
Before a cloud run starts, Writ checks your prepaid $ wallet and places a credit-hold for the estimated compute. If your balance can't cover it, the run is stopped with a clear 402 before any charge - so a run never quietly starts a bill you didn't agree to. A run on your own machine skips all of this: it is free.
Top up the wallet
Add funds in $5-$1,000 amounts. Purchased funds never expire.
Install once, then call it as your own API
A template gives you reusable workflow logic without another party's credentials or data. Attach your own account, test the operation, and publish it as REST or MCP.
| Local execution | Unlimited on your own machine. |
| Managed cloud | Your monthly pool of included usage first, then your top-up ceiling. |
Liveness and reviews tell you what to expect
Before you install, check the listing's liveness status - it shows whether recent real runs are succeeding and when the last successful run happened - and read reviews, which come only from people with at least three successful runs. Performance metrics are derived from real cross-tenant runs and stay marked provisional until the listing has 20 of them.
Observed, not re-tested. Liveness is a passive read of real activity. Writ does not run background re-tests on your behalf, and no money or compute is spent to produce the status.
If something goes wrong
Report a problem on any run and admins review it. When a creator edits a recipe, your install keeps running the version you tested and shows an "update available" notice - you pull the change only when you consent, never silently.
Install FAQ
Do I get the creator's logins?+
No. The creator's credentials, sessions, and secrets are stripped at publish. You attach your own persona or login and your own vault secrets, and the run resolves only your data.
What if the value I enter is wrong?+
A creator can attach a regex rule to any input or secret slot. Your value must fully match it or the run won't start - which catches a malformed value before the run begins.
What does a run cost?+
The listing costs nothing: every listing is free to install and free to run. A run on your own machine is free and unmetered. A run on the managed cloud is metered by actual running time, whether it succeeds or fails, because the cloud did the work.
Can my AI agent call it?+
Yes. Every installed listing is both a REST endpoint at /v1/{slug}/{path} and an MCP tool, authenticated with your own wt_ keys, so your software and your AI agents can run it on demand.
Can I run it offline or on internal sites?+
Yes. Run it on a local agent that reaches your own network - intranet, localhost, and VPN-only systems the cloud can't see - free and unmetered.
What's the liveness badge?+
A read of whether recent real runs are succeeding and when the last successful run happened. It is observed from real activity, not an independent re-test.
Ready to install your first one?
Browse the Marketplace, attach your own data, and run it on your account.