Session vault for autonomous agents
Automate the websites that refuse to have an API.
Most of the world’s real work happens behind a login, on portals that will never ship an API. PreLogin captures a genuine signed‑in session once, seals it in an encrypted vault, and lets your agents run the workflow nightly, in parallel, without ever touching a password again.
- No password storage
- Envelope‑encrypted at rest
- Per‑run browser isolation
- Append‑only audit trail
-
Capture
You sign in by hand, once, in a real browser. MFA, SSO, CAPTCHA, all of it, because it’s actually you.
-
Vault
The resulting session is sealed per tenant. Envelope‑encrypted, AAD‑bound, revocable. No plaintext credentials exist.
-
Run
Every run replays the session into a fresh, isolated browser context. The agent works as you: unattended, on a schedule.
# 1. Open a real browser; you complete the login once. curl -X POST $PRELOGIN/v1/captures \ -H "Authorization: Bearer $PRELOGIN_KEY" \ -d '{"startUrl":"https://portal.example.gov/login"}' # 2. Seal it, then let the agent do the work — every night, no password. curl -X POST $PRELOGIN/v1/runs \ -H "Authorization: Bearer $PRELOGIN_KEY" \ -d '{"sessionId":"pls_4f1c…","steps":[...]}'
The API was never coming.
Integrations get funded for the customers who ask loudly. Everyone else gets a login page, a CSV export that breaks monthly, and a support address.
-
Scrapers die at the login wall
Anything that fetches HTML anonymously stops dead at SSO, MFA, device fingerprinting and bot detection. That barrier sits right where the valuable data lives.
-
RPA breaks on the first redesign
Selector‑based automation is brittle by construction. One marketing change and your nightly job fails silently at 3am.
-
Credential replay gets you blocked
Typing a stored password into a login form trips every anomaly detector the site has. It also means holding plaintext passwords, the last thing any finance team wants on the books.
So we stopped fighting the login. PreLogin doesn’t impersonate you. It reuses the session you already established.
Session replay, not credential replay.
The distinction sounds academic. It is the entire product. Everything below follows from it.
-
01
MFA, CAPTCHA and SSO stop mattering
A human solves the challenge once, by hand. Afterwards the agent replays an already‑authenticated session, so there is no second factor to satisfy and no bot check to trip.
-
02
Cross‑origin iframes are actually usable
Login is almost always a cross‑origin iframe: Google Sign‑In, Apple ID, Auth0, reCAPTCHA, 3‑D Secure, Stripe Elements. Most tooling cannot see inside those frames at all. PreLogin reads, types and clicks into them.
-
03
Every run gets a clean room
Sessions replay into a fresh, isolated browser context per run. Two concurrent agents on the same portal cannot see or contaminate each other.
-
04
Nothing to leak
There is no password in the vault, only the session it produced. A breach yields cookies you can revoke in one call, not a credential that follows your customer for life.
-
05
Steps are declared, not scraped
Runs are expressed as typed step specs against a schema, so failures are explicit and reviewable rather than a silent timeout.
-
06
Every action is attributable
An append‑only audit log records who ran what, against which session, and when. That is the difference between automation you can defend and automation you have to apologise for.
Anywhere work is gated behind a login.
This is not a tax tool or a sales tool. It is the layer underneath, for every site you depend on that never gave you an API.
Tax & statutory portals
GST, income tax, MCA, EPFO. Monthly filings that are pure repetition.
Internal tools behind SSO
Admin panels and back‑offices nobody will ever expose programmatically.
Vendor & supplier portals
Pull invoices, push POs, reconcile statements across dozens of logins.
Banking & reconciliation
Statement downloads, payment status, multi‑account sweeps.
Insurance & claims
Claim status, document submission, broker dashboards.
Healthcare payer portals
Eligibility checks and remittance advice, without a clearinghouse.
Government filings & licences
Renewals, attestations, status checks on public portals.
Lead & market research
Data that exists only for logged‑in members of an association or exchange.
E‑commerce seller dashboards
Settlement reports, inventory, returns, per marketplace, per region.
Legacy back offices
Green screens and intranet apps that predate the concept of an API.
Built like the thing holding your credentials matters.
Because it does. The vault is the most sensitive component in the system, so it is designed on the assumption that everything around it fails.
-
01
Envelope encryption per tenant
Each tenant gets its own data encryption key, wrapped by a master key and bound by additional authenticated data. Rotating one tenant’s key never touches another’s.
-
02
No plaintext passwords, ever
The login happens in a browser you control. We only ever hold what the site handed back: a session, not a secret you can reuse elsewhere.
-
03
Secrets stay in your cloud
The master key and admin token live in your own secret manager and are read at boot. They are never baked into the image or passed as instance metadata.
-
04
Revoke and rotate on demand
Kill a single session, or rotate a tenant’s key, in one call. Both are recorded in the audit log.
Priced per workload, not per seat.
Start with one portal. Move the rest when it holds up.
-
Pilot
$490 / month
- 1 captured session
- 1 automated workflow
- Up to 500 runs / month
- Audit log, retained indefinitely
- Email support
-
Scale
$1,890 / month
- Up to 10 captured sessions
- Unlimited workflows
- Up to 10,000 runs / month
- Everything in Pilot
- Priority support, 1 business day
-
Enterprise
Custom
- Self‑hosted or dedicated VPC
- Unlimited sessions & runs
- Your KMS for the master key
- DPA & custom step ops
- Named engineer
The things people ask first.
Do you store my password?
No. You complete the login yourself in a real browser. We capture only the session the site returns, seal it, and discard everything else. There is no password in the vault to leak, because we never receive one.
How does this handle multi-factor authentication?
By not having to. A human satisfies MFA once during capture. Every subsequent run replays the already-authenticated session, so there is no second factor to satisfy on an unattended run.
Isn’t this just scraping?
Technically it is automation, but the posture is different: the agent acts as an identity the site already authenticated, on a session you explicitly created, with every action audited. It respects the site’s own access controls rather than trying to route around them.
What breaks it?
Substantial redesigns of the target site. Anything that invalidates sessions (a forced password change, an IdP session policy) requires a fresh capture. Because runs are declared step specs rather than brittle selectors, routine layout changes are usually absorbed.
Can we self-host?
Yes, on the Enterprise tier. The deployment is a single container plus a persistent volume, with secrets supplied by your own secret manager. Nothing about the hosted architecture is proprietary to our infrastructure.
How is this different from RPA?
Classic RPA automates the interface and stores credentials, which makes it brittle and a liability. PreLogin automates the session instead: no stored passwords, isolated browser contexts per run, typed step specs, and an audit trail designed for scrutiny.
Early access
Bring us the portal nobody could integrate.
Tell us which login‑gated workflow is eating your week. We’ll show you it running unattended, on your site, not a demo.