Layr DocsData and privacy
How Layr handles product data — access, AI processing, encryption, retention, and deletion controls that exist today.
Who this is for
Product conversations live in Slack, tickets, docs, and customer channels. That material is confidential.
If you are evaluating Layr for real project data, this page explains — in concrete terms — what we access, how we process it, how we protect it, how long we keep it, and which deletion controls exist in product today.
For a one-page buyer brief, start with the Security FAQ.
Share this with founders, product leaders, legal, and security reviewers.
Our stance
We treat your product data as customer data, not training fuel.
We design Layr so you start narrow, expand deliberately, and keep humans in control of anything that writes back into your tools.
We support enterprise review — including data processing agreement (DPA) discussion — as you move from evaluation into production.
Layr’s job is to help your team see recurring problems and decide what to build next, with evidence you can verify.
Review by topic
Use the tabs below to jump to the questions buyers usually ask first. The full detail follows on this page.
We store content you include, product memory we derive, search indexes, and account/security data needed to run your workspace.
Connecting a tool is not the same as ingesting everything. Sources stay excluded until someone confirms them.
We do not use your product data to train shared Layr models.
Model calls run on Layr’s backend. Your browser talks to Layr; Layr calls the model provider.
Embeddings, classification, answers, and related generation run server-side.
Customers do not send product content from the browser directly to a model API.
We encrypt integration OAuth tokens before storage, hash account passwords, verify inbound webhooks, and scope product and search access by workspace.
Disconnect, source exclusion, member removal, and account deactivation are available in product. Full workspace or account erasure is handled on request through privacy@uselayr.com.
We walk security teams through the controls we run and support DPA discussion for qualifying customers.
Formal audit reports, when available, are shared under NDA.
What Layr stores
When you connect tools and include sources, Layr may store the following.
We use this data to operate your workspace.
We do not use your product data to train shared Layr models.
How your data enters Layr
Connecting a tool is not the same as ingesting everything.
You authorize the connection
You complete the provider’s OAuth flow. Layr stores the credentials required to access what you authorized — not your password to that tool.
We discover sources; you confirm them
Layr lists available sources such as channels, projects, pages, and inboxes.
Sources remain excluded until an authorized user confirms inclusion.
Unconfirmed sources are not ingested. Most teams should start with a small set of high-signal places, not the entire company graph on day one.
We import and keep those sources current
After confirmation:
- We run an initial import of recent history for included sources
- We continue with ongoing sync and provider events where available
- We reject payloads that fail verification or fall outside your included sources
We normalize and build product memory
Included content is cleaned into a consistent form, attached to your workspace, and used to build the evidence and opportunity experience your team sees.
You control the perimeter: which tools are connected, and which sources inside those tools are in scope.
Integration permissions
We request least-privilege access for the workflows you enable.
Reading for analysis is the default.
Writing into your tools is a separate, explicit action.
| System | What we typically read when included | What write means in Layr |
|---|---|---|
| Slack | Channel and message history plus related metadata for included channels (public or private when the bot can access them). Direct messages are not in the current ingest path. | Optional interactive features only if enabled; not required for basic ingest |
| Jira | Issues and comments for included projects | Creating tickets only when someone intentionally publishes from Layr |
| Linear | Issues and comments for included teams or projects | Creating issues only on intentional publish |
| Notion | Pages and databases you share and include | Publishing a page only on intentional export |
| Intercom | Conversations you include | Read path for product signal |
Recommendations and drafts link back to original source evidence so your team can open the underlying message, ticket, or page before acting.
Full OAuth scope matrices are available during security review. See also Permissions and access.
How AI is used
Some Layr features use AI, for example:
- Turning text into searchable meaning (embeddings)
- Grouping related evidence into problems (classification)
- Answering questions over your product memory
- Helping draft specs from an opportunity
Call path
Model calls run on Layr’s backend. Your browser talks to Layr. Layr calls the model provider. Customers do not send product content from the browser directly to a model API. A network inspector on a user’s laptop shows traffic to Layr, not a direct browser-to-model session.
Provider
Model inference for these features runs on Layr’s servers through contracted model providers.
We do not train shared Layr models on your product data.
Provider handling of inference content follows that provider’s terms and the configuration we run for customer workloads.
For procurement we can discuss:
- Which model families are enabled for your environment
- How prompts and outputs are handled
- Subprocessor listing and DPA alignment
Where your data lives
Layr keeps product data in our production application infrastructure:
| Layer | Role |
|---|---|
| Primary database | Durable system of record for accounts, workspace data, evidence, opportunities, and credential metadata |
| Queue and cache layer | Reliable background jobs and operational state |
| Vector index | Meaning-based search over product memory |
| Object storage | File objects when a feature needs them |
Marketing at uselayr.com and documentation at docs.uselayr.com are hosted separately from the product data plane.
Region, architecture diagrams, and subprocessor details for your contract are shared in security review or under NDA as needed.
Encryption and protection
Concrete controls we run today:
Retention
We retain workspace data so the product keeps working for your team.
Evidence history, opportunities, and search memory remain available while the workspace is active and sources remain connected under your settings.
Initial sync lookback — how far back a first import goes — is not an automatic deletion timer.
If your pilot or contract needs a fixed retention schedule, we set that in the agreement and operate to it.
See also Data retention.
Deletion, disconnect, and offboarding
This is the section security reviewers care about most. Here is the concrete model.
What you can do in the product:
- Disconnect an integration — Layr stops using that connection for further ingest from that tool
- Remove sources from inclusion — newly excluded sources stop contributing new evidence
- Deactivate a user account — sessions are revoked; that user loses normal product access
- Remove workspace members — membership access ends for that person
Full workspace or account erasure is handled on request.
When an authorized admin contacts our privacy channel, we confirm scope, execute the agreed removal process, and communicate what is removed. Legal or billing records we are required to keep may remain in minimized form where the law or payment rules require it.
- An authorized customer request (owner, designated admin, or contract process)
- We confirm scope (which workspace or account)
- We complete the agreed removal process
- We confirm completion of that scope
Contact privacy@uselayr.com.
What you control every day
| Control | Meaning |
|---|---|
| Connections | Which tools are authorized |
| Source selection | Which channels, projects, or pages are in scope |
| Human approval for writes | Nothing important is silently written back to Jira, Linear, or Notion without intent |
| Roles | Who can connect, review, and administer |
| Opportunity review | Your team decides what becomes trusted product direction |
Layr is built to earn trust before it asks for deeper access.
Enterprise review
We support the conversations enterprises expect:
- Clear processing purpose — provide Layr to your workspace
- Minimization via source selection and least-privilege scopes
- Security measures described on this page
- Subprocessor transparency, including server-side AI inference
- DPA discussion for qualifying customers
- In-product deletion controls plus erasure handling on request
When a formal audit report is available, we share it under NDA. Detailed control inventories and questionnaires are available through privacy@uselayr.com.
How we recommend starting with real data
Connect one primary tool
Include a small set of high-signal sources only
Keep write-back gated until the workflow feels right
Invite security early if you need DPA or subprocessor review
Expand source coverage after the first slice looks correct
Contact
For security review, DPA, subprocessors, or erasure:
Closing
Layr is built for teams who cannot be casual with product data.
We use server-side AI through contracted providers, encrypt integration credentials before storage, scope access by workspace, and keep humans in control of source selection and write-back — with erasure handled on request through our privacy channel.
If that matches how your team evaluates vendors, we are ready to go deeper with you.