Built for the documents you’d never email.

~200

customer tables covered by Postgres row-level security

24

automated cross-tenant isolation suites run in CI on every change

60 s

signed download URL lifetime, minted per request, never pre-signed into HTML

40 MB

upload cap, with the file's actual byte signature checked, not its extension

12+

password minimum, screened against known breach corpora before acceptance

1 year

maximum age of API keys and SCIM tokens, stored as SHA-256 hashes

15 min

audit event batches streamed to your SIEM, signed and cursor-tracked

3 days

business-day SLA to acknowledge a vulnerability report, with safe harbor

Open your Security Center

In the app under Settings, Security, Posture: your org's MFA coverage, audit export, and SIEM webhook setup, live, on your own tenant's data.

Run the isolation test

Create two organizations, upload a document to one, and confirm the other's session cannot read it. The denial comes from the database's row-level security, not an application error page.

Watch your own audit stream

Subscribe an audit webhook and signed, cursor-tracked batches of your org's events arrive within 15 minutes. Compare them against what your team actually did.

Test deletion end to end

Delete a contract and confirm rows and files are gone, then download the timestamped deletion certificate that the deletion wrote.

Row-level tenant isolation

Every one of roughly 200 customer tables carries your organization ID, and Postgres row-level security policies enforce the boundary on every query. 24 automated cross-tenant isolation suites run in CI on every change: the failing cross-org read is a permanent regression test, not a demo.

Private file storage

Contracts live in private buckets with storage policies mirroring the same org rule. There is no public URL to any customer file, ever.

60-second signed downloads

Every file download uses a signed URL minted per authorized request and expiring in about a minute. Nothing is pre-signed into a page, so a leaked link or a saved HTML file is dead before it can be abused.

Encrypted everywhere

TLS for everything in transit; database and storage encrypted at rest. No secret is ever stored in the codebase, keys live in server-side environment variables only.

Hardened headers

HSTS, nosniff, frame-deny, and a strict Content-Security-Policy carrying a fresh per-request nonce, which is why every page renders dynamically instead of from a cache that could serve a stale policy.

Rate limiting

Authentication, upload, and AI routes are rate-limited per user and per IP, with per-key limits on the public API. In production the limiter fails closed.

TOTP multi-factor auth

App-based MFA for all users with session step-up enforcement, enforceable org-wide by policy, and mandatory for every member of our own staff.

SCIM 2.0 provisioning

Okta, Microsoft Entra, or any SCIM-capable IdP provisions and deactivates users automatically, offboarding in your directory is offboarding here.

Role-based access

Owner and member roles per organization, managed by your own org owners. Sensitive actions like member management and deletions are owner-only.

Admin rights can't be self-assigned

Platform staff access lives in a separate table that no customer-facing code path can write to. There is no privilege-escalation route from a customer account.

Per-contract access control

Restrict individual contracts to named people within your own organization, need-to-know inside the tenant boundary too.

Hashed, expiring API credentials

API keys and SCIM tokens are stored as SHA-256 hashes and expire within a year. We can verify them; nobody can read them back.

Enforced SSO

Switch on Require SSO and password and magic-link sign-in are refused for your domain. Owners keep a break-glass path so a broken IdP cannot lock you out.

Session control

See every device your account is signed in on, revoke any one, or sign out everywhere else. A sign-in from a new device emails you the same minute.

Breached-password screening

New passwords need 12 characters and are checked against known breach corpora with a privacy-preserving query before they are accepted.

Upload validation

Allowed document types only, a 40 MB cap, and a file-signature check on the actual bytes, a renamed executable doesn't get in because the extension says PDF.

Full audit trail

Every view, download, create, update, and delete of contract data is logged with actor, action, entity, organization, IP, and user agent. Analyst access to your files is logged the same way, in the trail you can inspect.

Real deletion

When you delete a contract, both the database records and the stored files are removed, and deleting your organization sweeps every storage bucket. Deletion is a data path, not a support ticket.

Confidentiality scanner

Anything shared outside your organization is scanned for sensitive terms first, and external share links expire.

AI without training

AI features run against your tenant-isolated data at request time. Your contracts are never used to train models, ours or anyone else's. The full data flow is spelled out below.

Anonymized benchmarks

Benchmark aggregates are pooled and anonymized. Nothing that could identify your company, contract, or terms is ever exposed to another customer.

Audit stream to your SIEM

Signed batches of audit events push to your SIEM every 15 minutes, cursor-tracked so nothing is skipped, and owners can export the complete history as CSV any time.

Automatic security alerts

An hourly detector reads the audit trail and emails your owners when one account's downloads, exports, or permission changes look unusual.

Encrypted connector credentials

SAM, spend, and workflow connector credentials are encrypted at rest, scoped to your organization, and used only server-side.

Verified inbound

Email-in and CLM webhooks verify HMAC signatures before anything is processed; each org's inbound addresses carry unguessable capability tokens.

Signed outbound webhooks

Events we send you are HMAC-signed with timestamped signatures, so your systems can verify us the same way we verify them.

Nothing leaves until you act

AI runs only when a member of your org runs an AI feature. There is no background training job reading your documents.

What leaves, and to whom

The specific document text and question needed for that one run go to Anthropic (Claude), our only model provider. Nothing goes to any other AI vendor, ad network, or analytics service.

What comes back, and where it lives

The analysis returns to your tenant and is stored under the same row-level security as the document it came from.

Retention and training

Anthropic does not use API data to train models. A contractual zero-data-retention addendum is in progress and tracked on the claim ladder above until it is signed.

Attribution

Every AI run is metered and attributable to the user and organization that ran it, so your admins can see who analyzed what.