This page is intended for customer security reviews, pilot conversations, and beta onboarding.
It explains how Teamli handles tenant isolation, authentication, document access, auditability,
and operational security today.
One EU region (europe-west1), all Google Cloud. Every request authenticates and is placed into your
company's database context before any data is read. The front-end holds no customer data.
Browser
React SPA
Served over HTTPS by Firebase Hosting (static + CDN). Holds no customer data.
→
API
Express API
Runs on Google Cloud Run. Verifies identity and enforces roles on every route.
→
Data
PostgreSQL
Google Cloud SQL with row-level security. Encrypted at rest with AES-256.
Cloud Storage · documents & photosAI assistant · no training on your dataCloud KMS · encryption keys (CMEK)Secret Manager · keys
The security-critical controls live in the platform, not our app code, so the
guarantees hold even if our application has a bug.
How your company's data stays isolated
Tenant isolation is enforced below the application, in the database, using
PostgreSQL FORCE row-level security.
1
Request arrives
An authenticated API call comes in with the user's sign-in token.
→
2
Identity verified
Firebase verifies the token; we resolve the user's company and role.
→
3
Tenant context set
The database session is locked to your company before any query runs.
→
4
Only your rows
FORCE RLS filters every query, so no result can include another company's data.
FORCE RLS applies even to the database role our app connects as, so
a buggy query still cannot cross companies.
Frequently Asked Questions
How is customer data separated between companies?
Teamli uses tenant-scoped data access enforced at the database layer with PostgreSQL
row-level security. API requests are placed into tenant context before tenant data is queried,
which reduces the risk of cross-tenant data exposure through application bugs alone.
How do users sign in?
Organizations can enable Google, Microsoft, and email/password sign-in. Identity is handled
through Firebase Authentication, and the application maps the signed-in user into the correct
organization and role before allowing access to tenant data.
1
Enter email
We look up which sign-in methods your company's domain allows.
→
2
Choose method
Google, Microsoft, or email/password, whatever your company enabled.
→
3
Identity verified
Firebase Authentication confirms who you are.
→
4
Mapped to your org
You're placed into the right company and role before any data loads.
Can your internal admins see our employee data?
Not by default. Platform-only super admins are separated from tenant-scoped workflows unless
they are deliberately operating with tenant context. The goal is to keep platform
administration and customer workforce access from being mixed casually.
Do we control when Teamli support can access our workspace?
Teamli staff have no standing access to your data. Operator access is time-boxed and
reason-required, and can require that every access request is approved by your own
administrators before it is granted, with an in-app history of when and why your workspace
was accessed (Settings → Security). Approving access to compensation data is restricted to the
account owner.
1
Support requests
A reason and a time limit are required: no request, no access.
→
2
You approve
Your admin reviews it in-app; compensation access is owner-only.
→
3
Time-boxed grant
Access auto-expires after the approved window, leaving no standing access.
→
4
Fully logged
Who, when, and why is recorded in your audit history.
Teamli · Settings · Security
Support access request
Review and decide before any access is granted.
TS
Teamli Supportrequesting temporary access
Reason
Investigating a reported export issue
Scope
Directory · no compensation
Expires
Automatically in 4 hours
ApproveDeny
Can we restrict access to our workspace by IP address?
Yes. Admins can configure an IP allowlist (IP ranges in CIDR notation) under
Settings → Security, so your workspace is only reachable from your approved networks.
Requests from any other IP are refused and receive no data. It is enforced by Teamli's API
and covers access through the app.
Teamli · Settings · Security
IP allowlist
Restrict access to approved networks
203.0.113.0/24Tel Aviv office×
198.51.100.10/32VPN gateway×
Requests from any other network are refused and receive no data.
How do you control what employees, managers, and admins can see?
Teamli uses role-based access control in the API and shapes the UI around those
permissions. Employees are restricted to self-service views, managers get direct-report access
where appropriate, and admins get organization-level operations. Sensitive fields such as
salary and some document classes are further filtered by access level.
Can we define our own roles and scope or time-box access?
Yes. Beyond the built-in levels (admin, manager, employee, with or without compensation
access), an administrator can create custom roles under Settings → Roles
and assign them to individual people, whole departments, or sites, either permanently or
for a fixed date range (a temporary grant that reverts automatically when it
ends). Assigning shows a before→after preview and a "who has it" view, and every role
change is recorded in your audit log. Access is enforced in the API and, for data, by the
database's row-level security; a person's role level and compensation-visibility scope govern
what they can see and do.
How are documents protected?
Documents are stored in Google Cloud Storage and accessed through signed URLs instead of
public file links. Access checks are done before the URL is issued. Document access is scoped
by employee, role, and, in some cases, document sensitivity.
How do document e-signatures work, and how are they protected?
Teamli supports e-signatures on PDF documents: offer letters, policies, contracts. An
authorized user sends the document (or a reusable template whose fields auto-fill from the
employee's record; a send is refused if required data is missing, so a contract never goes
out with blanks); signing happens through a dedicated e-signature provider; unsigned signers
are reminded automatically and requests expire past their deadline; and the completed, signed
PDF is stored back in your workspace with a full audit trail and a downloadable
completion certificate (who signed, and when; available only to people
authorized to see the signed document itself).
1
Send for signature
An authorized user picks a PDF or a template and the signer(s) in Teamli (PDF only, up to 4 signers; templates support sequential signing).
→
2
Signer signs
Each signer gets a secure link and signs in the provider's hosted flow.
→
3
Completion verified
The provider notifies Teamli; the callback is cryptographically verified before it's trusted.
→
4
Stored & sealed
The signed PDF is saved in your workspace and the whole sequence is audited.
How it's protected
AccessOnly authorized roles can send, view, or download a document, and compensation documents stay hidden from anyone without salary-visibility permission.
AuthenticityCompletion callbacks are verified with an HMAC-SHA256 signature (timing-safe), so forged or unsigned callbacks are rejected, so a signature can't be faked into your records.
StorageThe signed PDF lives in Google Cloud Storage, encrypted at rest, reached only through short-lived signed URLs, never public links.
Audit trailEvery step (sent, resent, revoked, downloaded, expired, certificate viewed, completed) is written to your tenant audit log; audit entries carry signer counts, not signer email addresses, so signer emails aren't duplicated into the log store.
Can documents be sent for signature automatically (workflows, bulk, AI)?
Yes, with deliberate guardrails, and a human decision wherever judgment is involved.
WorkflowsA workflow (e.g. new-hire onboarding) can send an e-sign template to the person it's about. Workflows are admin-authored, every automated send is audited with the rule that triggered it, and sends carry a database-enforced idempotency key: a replayed or retried event finds the earlier send and stops instead of sending again.
Bulk sendAn admin can send one template to a filtered audience, with the exact recipient count shown before confirming, per-recipient error isolation, and the same per-recipient idempotency key (re-running a batch skips everyone already sent). People who already have the same document in flight are skipped.
AI renewalsAn optional, admin-enabled agent watches signed contracts nearing expiry and proposes re-sending the template. Nothing is sent without an explicit admin approval: the proposal waits in the Approvals inbox, is re-validated at the moment of approval, and executes through the same audited, role-checked action layer as every other AI action. AI features are off by default and require both platform-level enablement and your organization's own opt-in.
CompensationSalary-bearing documents are excluded from every automated path: they can only be sent manually, one at a time, by someone whose salary visibility over that specific employee is verified at send time. Automation can't become a way to route salary data around the permission model.
Do you keep audit logs?
Yes. Teamli records audit events for sensitive actions such as employee updates, tenant
administration, document actions, lifecycle workflow actions, and other privileged operations.
Admins can review the log in-app and drill into any event to see exactly what changed
(a before → after view), who did it, and when. They can filter by actor, date, or a specific
record (that record's full history), and export to CSV. Compensation values in audit detail are
shown only to admins with salary-visibility permission. Auditability is a core part of the design.
Teamli · Audit log · Event details
Event details
Action
employee.updated
Record
employee · EMP-1042
Actor
hr@contoso.com
When
Jun 29, 2026, 14:32
Changes
jobTitle
Analyst→Senior Analyst
salary
[redacted: requires salary-visibility permission]
Do you support single sign-on?
Teamli supports Google and Microsoft sign-in through Firebase Authentication, plus
email/password where needed. The login experience can be configured per organization so users
are guided to the right provider for their company.
Do you log sign-ins and sensitive data reads?
Yes. Every sign-in is recorded in the tenant audit log (provider and source IP), and sensitive
reads are logged, not just writes: viewing someone's salary, opening their compensation history,
and exporting data all leave audit entries. On top of the log, anomaly tripwires alert our
operations channel in real time: bursts of failed authentication from one source, and any
privilege elevation such as an admin-level or compensation-bearing role being granted.
How long do sessions last, and can access be cut off instantly?
Access tokens are short-lived (about an hour) and every API request re-checks revocation, so
disabling a person or changing their permissions takes effect immediately, not at the next
login. Sessions also have an absolute lifetime (30 days by default): after that, a fresh sign-in
is required, so a stolen "remember me" token can't live forever. Changing someone's compensation
visibility revokes their active sessions on the spot.
Do you support multi-factor authentication (MFA)?
Yes, and for administrators it is required, not optional. An administrator signing in with
email/password must hold a second factor (a TOTP authenticator app such as Google Authenticator,
1Password, or Authy); an admin sign-in without one is refused and routed into a
one-time enrollment. Administrators who sign in with Google or Microsoft are covered by that
provider's own MFA. This is built on Google's managed authentication platform rather than a
home-grown implementation.
Teamli · Security · Two-factor authentication
Set up two-factor authentication
Required for administrators.
Scan this code with your authenticator app.
Enter the 6-digit code it shows.
419027
Verify & enable
How do you protect against common abuse patterns?
The API uses production CORS restrictions, rate limits, database query timeouts, and
sanitized error responses. Destructive actions use deliberate confirmation flows rather than
one-click deletion.
Where is the system hosted?
Teamli runs on Google Cloud managed services. The frontend is delivered through Firebase
Hosting, and the API and data services run on GCP-managed infrastructure.
Is data encrypted?
Yes. Data is transmitted over HTTPS/TLS, and managed GCP services provide encryption at rest
for stored data.
How is data encrypted at rest, including salary fields?
All data at rest (the entire database on Google Cloud SQL, including every employee and
salary field) is encrypted with AES-256 under customer-managed encryption keys (CMEK)
held in Teamli's own Cloud KMS. Files are encrypted the same way, under a dedicated
storage key. (For who controls the keys, and the limits of each option, see
"Who controls the encryption keys?" below.) Compensation data is
further protected at the application layer by database row-level security and role-based
permissions, so only authorized users can read it, and it is never exposed through the
external API.
In transitHTTPS / TLS on every connection: browser to API, API to database.
At restAES-256 across the whole Cloud SQL database and Cloud Storage files, with CMEK keys in Teamli's Cloud KMS, rotated every 90 days.
AccessRLS + role permissions gate compensation, and it is never returned by the external API.
Who controls the encryption keys?
Today: Teamli-managed keys (CMEK), AES-256 across the database, files, and backups, under
keys held in Teamli's own Cloud KMS. Key control comes in three tiers:
(1) Teamli-managed keys (CMEK, live today). Key lifecycle is under our control
in a dedicated key ring: automatic 90-day rotation, key usage visible in our audit logs,
alerting on key-state changes, and a platform-wide kill switch; this is vendor governance,
not customer key control.
(2) Customer-held keys for pay data (roadmap). Compensation fields are
envelope-encrypted with a per-customer key that can live in the customer's own cloud project, so
revoking it renders that customer's pay data unreadable, including in backups.
(3) Teamli Private (BYOC). The entire deployment runs in the customer's own
cloud project, where their key wraps the whole database and storage: the only architecture where
"you hold the key to your data" is literally true.
What is your security philosophy in one line?
We do not invent our own security primitives. The security-critical controls (identity and
tenant isolation) are enforced by Google's managed platform and by PostgreSQL itself, not by
our application code, so the guarantees hold even if our application has a bug. Google secures
the platform; we configure it without weakening it. The three foundations are Firebase
Authentication, Google Cloud, and PostgreSQL row-level security.
Does tenant isolation depend on your application being bug-free?
No. We use forced row-level security, which applies even to the database role the application
connects as, so no query, even a buggy one, can return another customer's data. Isolation is
enforced below the application, in the database.
Is our data used to train AI models?
No. Your prompts and data are processed per request, statelessly, and are not used to train
any model. Model inference runs under a data-processing agreement with zero data retention.
Does the AI use a separate instance per company?
It does not need to. The model is a shared, managed service, and per-tenant instances
are not required for isolation. What keeps companies separate: the data placed into each prompt
is already scoped to the asking company by row-level security; inference is stateless, so
requests cannot bleed into each other; and your data is not used for training. The assistant
is also bound by the same access rules as the person asking, so it cannot surface anything they
could not already see.
Do you have an API, and what can it read?
Yes: scoped, read-only service-account API keys. They can read people (the directory), org
structure and departments, and approved time-off; compensation is never exposed. Keys are
hashed at rest, shown only once, revocable, expirable, and audited, and they use a separate API
surface that cannot reach the internal application routes. External systems can also receive
signed webhooks.
Do you support SCIM / Microsoft Entra provisioning?
Yes, SCIM 2.0 with Microsoft Entra ID. Your directory can auto-create, update, and deactivate
people in Teamli, so people who leave the directory are removed automatically.
What is the “SPA / Firebase Hosting” on the architecture diagram?
The SPA (single-page application) is our React front-end, the interface that runs in the
browser. Firebase Hosting is Google's static hosting and CDN that serves it over HTTPS. The
front-end holds no customer data; everything flows through the API, where authentication and
tenant isolation are enforced. The front-end being public (like any website's code) is not a
risk: nothing is shown until the user authenticates and the API returns only their company's
data.
How do you ship changes safely?
Every change is peer-reviewed and gated by 1,100+ automated tests, including
tenant-scoping tests, so isolation cannot silently regress. Every pull request also passes
security gates in CI: static analysis, secret scanning over the full git history, and a
dependency audit that fails the build on high-severity advisories in production dependencies,
plus a weekly scheduled sweep. CI actions are pinned to exact commit hashes, and newly published
package versions wait out a 7-day cooling-off period before we install them. Database migrations
are numbered and append-only, and Cloud SQL provides backups with uptime and latency alerts.
What is your backup and disaster-recovery policy?
Teamli runs on Google Cloud SQL for PostgreSQL (managed). The database is backed up automatically
every night, the last seven backups are kept, and every backup is encrypted at rest. A restore
returns the database to the most recent successful backup, so changes made since that backup could
be lost. Documents and photos are held in Google Cloud Storage in our EU region, where Google keeps
copies across separate zones. The application and API are stateless, so recovery is restoring the
database and redeploying. We ran our first restore drill on 6 October 2026, on our development
environment: a backup was restored into a separate database and checked against the original.
Creating the database, restoring and checking took under half an hour, not counting reconnecting the
application. A drill on production comes next, and we don't yet publish committed recovery-time or
recovery-point targets.
Nightly
Automated backups
Cloud SQL backs up the database every night, encrypted at rest.
→
7 backups
Seven kept
The last seven successful backups are kept to restore from.
→
Recover
Restore & redeploy
The app is stateless, so recovery is restore the database and redeploy.
Committed recovery-time / recovery-point targets and a restore drill on production are on the
beta hardening roadmap. The first drill ran on our development environment.
Are you SOC 2 certified?
Not yet. Teamli is in beta, and we've focused first on building the security foundation
(tenant isolation, encryption, access controls, and audit logging) rather than formal
attestations. If SOC 2 or another framework is a requirement for you, tell us, and we'll
walk you through our timeline and the evidence we can share in the meantime.
What happens if there is a security issue?
We follow a written incident response procedure. It names an incident lead and a deputy, sorts
incidents into three severity levels, and sets the steps: preserve the evidence, contain the harm,
investigate, fix and recover. We notify affected customers within 24 hours of becoming aware of an
incident that concerns their data, and keep them updated until it is contained. Every incident is
recorded, and a serious one gets a written review. The procedure calls for a yearly exercise and a
yearly restore drill.
Can we review security before joining the beta?
Yes. You can review our architecture summary, auth and access model, data-isolation
approach, audit and admin controls, and the current beta hardening roadmap. Just ask your
Teamli contact.
How would you sum up your security posture?
Security is an active operating discipline for us, not a checkbox. We're happy to walk you
through the controls, the architecture, the auditability, and the areas we're still
hardening during beta.
We don't ask you to trust a slogan. We'll show you the controls, the auditability, the
isolation model, and the process we use to operate the system responsibly during beta.
Security Review Contact
For pilots and security reviews
Contact your Teamli beta point of contact to request a live walkthrough of architecture,
access controls, audit trails, and the beta hardening roadmap.
We can walk customers through the security overview, auth and tenant-isolation model,
document handling approach, and the current beta operations checklist.