SourceLace Docs
FAQOpen the app

Security and privacy

What SourceLace stores, what it never stores, how it protects what it keeps, and where your data goes. Written for your security reviewer: every statement here describes how SourceLace works today. Questions: security@sourcelace.com.

The short version

  • Each person uses their own login to most systems, so people only ever see and change what they already could. A few systems use one shared account set up by your admin instead, and SourceLace labels them everywhere; see Personal sign-in and shared accounts.
  • Read-only by default. Changes are off for every source until an admin turns them on for named objects, and even then every change is previewed and only sent after the person confirms it.
  • Your records are not copied. Query results are held for at most 30 minutes, in server memory or encrypted in a short-term store, and are never written to a database. Your admin can make this shorter, per source too, and have them deleted as soon as a chat is closed. There are two exceptions, both encrypted and deleted after your retention period: saved chats in the SourceLace app, and AI agents' run summaries and the changes waiting for approval (which hold the before and after values). Every answer is read live from your systems, unless your organization has chosen Saved answers, an optional add-on your admin turns on system by system.
  • Secrets are encrypted with a key for your organization alone, and never shown, returned or logged.
  • Everything is audited: every call, by every person and every client, in a tamper-evident trail that holds metadata, not your data.

Filling in a security questionnaire? Security questions answers the usual questions one by one, including what is not in place yet. For short answers by topic, see the FAQ on Security and data protection and Audit and logging.

What SourceLace stores

What Where and how
Your organization, its people and their roles, groups, sources and their options (and when an admin's connection test of a source last failed), skills and projects, and who each skill and project is shared with SourceLace's database.
Each person's sign-ins to your systems (access and refresh tokens), database passwords, Oracle and Workday client secrets, custom connectors' shared secrets, single sign-on client secrets, your own AI key SourceLace's database, encrypted with your organization's own key (see Encryption). Never shown again after saving, never returned by any API, never logged.
Saved chats in the SourceLace app: messages, the tables, charts and tool results in them, files made in them, and photos attached to questions (only if your admin turned photos on) SourceLace's database (files made in chats in SourceLace's private file storage), encrypted with your organization's key, and deleted after your retention period (30 days after last use by default; admins choose up to the longest your plan allows). People can delete their own chats at any time. When a pilot ends, nothing is deleted. If your pilot ends without a contract and SourceLace closes your organization, your saved chats are deleted 30 days after that.
AI agents: their instructions and settings, each run's times, counts and steps (never row values), and each run's summary, what triggered it, and changes waiting for approval with the record's current and proposed values SourceLace's database. Settings, summaries, triggers and pending changes are encrypted with your organization's key. An email that starts an agent is kept only as what triggered its run (subject, sender, date and text), and its readable attachments, encrypted, with the run; an agent's list of allowed senders is encrypted too. Runs and approvals are deleted with the agent and after your retention period; a change waiting for approval expires after 7 days. See AI agents.
Matching rules: how records of two of your systems are recognised as the same (object and field names, comparisons), their versions, editors, and the counts of each test, never a record, a key or a matched pair SourceLace's database. Rule names, ways and notes are encrypted with your organization's key. Deleted with the rule. See Matching records across systems.
Each person's profile: the name they chose, their theme and text size, and their picture, if they added one (their own upload, or the one their Google account shows, copied only when they choose Use my Google picture; reduced to a small square, so nothing else from the file is kept) SourceLace's database. The picture is encrypted with your organization's key. Deleted when the person removes the picture or is removed from your organization. See Your settings.
Your plan and billing: plan, seats, discounts, and Stripe customer and subscription ids SourceLace's database. Card details are held by Stripe, never by SourceLace.
Your account's records for the people who run SourceLace: the name, email address and phone number of your admin contact, your contract's start and end dates, and notes about the contract SourceLace's database, encrypted with SourceLace's own key. Only SourceLace's operators see them; nobody in your organization does. Changes to the contact are recorded in your audit trail as "changed", never with the values.
Favourite folders in Amazon S3, Azure Storage and Google Cloud Storage sources: folder paths only, never file names or contents The organization's are an option of the source. Each person's own (up to 20 per source) are in SourceLace's database, encrypted with your organization's key, and deleted with the source.
The audit trail SourceLace's database. Metadata only (see The audit trail).
AI usage One row per question: who, when, which model and token counts. Never the question or the answer.
Activity counts How many people were active, how many questions were asked, agent runs and similar counts per organization and day, which SourceLace uses to run and improve the service. Never your data: never a question, an answer, a record or a value, and people only as a one-way code, never their email address. Kept in SourceLace's own database and never sent to an outside analytics or tracking service.
Sign-in codes, one-time links and tokens SourceLace issues to MCP clients Stored only as SHA-256 hashes, so a copy of the database cannot be replayed as tokens.

What SourceLace does not store

  • Rows from your systems. Query results are held, for the person who asked, for at most 30 minutes (up to 2,000 rows a query), then forgotten: in server memory or, where SourceLace is set up with one, encrypted with your organization's key in a short-term store. They are never written to a database or a log, outside saved chats and AI agents' run summaries and pending changes. A larger answer (up to 250,000 rows) is kept whole for the same time in a short-lived file encrypted with a key of its own, which is itself kept only encrypted with your organization's key; the file is stored only in that encrypted form, so the answer stays available for follow-up questions during that time, and is deleted when the result expires or the chat's results are deleted. Your admin can make this stricter on the Data retention page (see Data retention controls). Every answer is read live from your systems, unless your organization has chosen Saved answers, an optional add-on your admin turns on system by system.
  • Files from your file stores. SourceLace reads a file only when someone asks, pulls its text out in memory, and does not crawl or index files. Every answer is read live from your systems, unless your organization has chosen Saved answers, an optional add-on your admin turns on system by system. A search inside the files of a bucket reads them the same way, for that one search. A Search index source sends searches to your organization's own index and keeps no copy of it.
  • An index of your systems. Searching everything asks each system live with its own search and the person's own access; SourceLace keeps no index or copy, and its results are held like any other answer.
  • Passwords to SourceLace. People sign in to SourceLace with Google, Microsoft or your single sign-on provider.

Data retention controls

Admins see every timer in one place, on the Data retention page, and can only make them shorter than SourceLace's limits, never longer:

  • How long query results are held: 1 minute up to 30. A query's full result is kept briefly so follow-up steps (joins, charts, documents) can use it, encrypted with your organization's key.
  • How many rows are kept per result: up to 2,000.
  • Per source: shorter times or fewer rows for a sensitive system, for example 2 minutes for an HR system and 30 for the CRM.
  • How long a previewed change can be applied: up to 10 minutes.
  • Where results are held: SourceLace's results store (shared by SourceLace's servers, encrypted with your key), or server memory only, never written to the shared store or kept anywhere outside server memory (larger answers included), and gone on any restart.
  • Saved answers (only if your organization has them): for each system, whether people's own recent answers may be reused and for how long, at most 24 hours; off unless your admin turns it on. See Saved answers.
  • Forget on close (on by default): a chat's results are deleted as soon as the person closes the chat, switches to another one or signs out, without waiting for the timer. Deleting a chat always deletes its results.

The page also shows the longer records: chat history (and files saved with chats), agent run records and agent approvals (7 days to decide).

Proof, not trust. Each result held writes an entry to your audit trail: which source, how many rows, how long and where it is held, and the time it is deleted (for example "held 30 min ... deleted at 14:32"). An early deletion (chat closed, chat deleted, signed out) writes its own entry with the reason. They hold no row values. Your compliance team can check them on the Audit log page.

Encryption

  • In transit: SourceLace is served over HTTPS only, with TLS 1.2 or later. Connections from SourceLace to your systems use TLS, and SourceLace checks the certificate; for databases it also checks the certificate was issued for the host name your admin entered, and TLS cannot be turned off for any database other than one on the same computer as SourceLace.
  • At rest: each organization has its own key. Secrets, saved chats and AI agents' settings, run summaries and pending changes are encrypted with it. Your organization's key is itself never stored in readable form, and the key that protects it is never kept in the database. Without your organization's key, nothing encrypted for your organization can be read. On top of that, everything SourceLace stores, backups included, is encrypted at rest by the hosting environment. Your admin can request the security review pack for the technical detail.

Keys: who brings each one, and where it is kept

Key Who brings it Where it is kept
Your organization's key Made automatically for your organization Stored only in protected form, never readable from the database alone. It locks everything below. Deleted for good when your organization is deleted.
Your own AI key (an Anthropic, OpenAI, Google Gemini, Mistral, Azure OpenAI or Amazon Bedrock API key) Your admin, only if you choose to use your own, in the app's settings. Otherwise SourceLace's own Anthropic key is used, included in your plan. The database, encrypted with your organization's key, never shown again. On a SourceLace Runner, it is kept in the runner's own database inside your network and never sent to SourceLace.
Each person's sign-ins to your systems (access and refresh tokens, database passwords) Each person, when they click Connect The database, encrypted with your organization's key.
Your own app secrets (single sign-on, Oracle, Workday, your own Slack app, SAP BTP, custom connectors, Amazon S3 access keys) Your admin, on the source or single sign-on page The database, encrypted with your organization's key, never shown again.
Tokens SourceLace gives you (SAP ECC agent, runner, AI app connections) SourceLace, shown to you once Only a fingerprint (hash) is kept, so a copy of the database cannot be used to sign in. Revoke one at any time.

Nothing in this table is ever returned by an API, written to a log, put in the audit trail or included in a support bundle.

Access control

  • Signing in to SourceLace: with Google, Microsoft, or your own single sign-on (OIDC or SAML 2.0, such as Okta or Microsoft Entra ID; SAML is new; see Single sign-on). Only people your admins add, or people from your organization's allowed email domains, can sign in. A session in the SourceLace app lasts 7 days.
  • MCP clients sign in through SourceLace with standard OAuth. After the person signs in, SourceLace asks them to allow the app, naming it and the address it will send them to, before anything goes back to it. Their access tokens last 1 hour. The refresh token lasts 30 days and is replaced each time it is used, so an AI app not used for 30 days signs in again. Removing a person (or the email domain they joined through), or revoking the app, ends both.
  • Roles: admins manage people, sources, groups, single sign-on and AI settings; members use the sources they may use. See the Admin guide.
  • Groups: admins can limit each source to named groups of people, and sync groups from your identity provider. Read rules narrow it to the rows each group may see, and the most rows one answer may return. See Access by group and Read rules.
  • Your systems' own permissions apply to every call to a source with a personal sign-in, because each person signs in with their own login. SourceLace adds its own limits on top, never removes any.

Personal sign-in and shared accounts

Every source is labelled, on Connections, when an admin adds one, on My connections, and for AI apps:

  • Personal sign-in: each person signs in with their own account, and the system applies its own permissions to each of them.
  • Shared account: everyone who may use the source uses one account, key or role that an admin set up. SourceLace's access by group, read rules and data protection rules still limit what each person sees, and the audit trail still records who asked, but the system itself sees the same account for everyone and cannot apply its own per-person rules.
Sign-in Systems
Personal sign-in Salesforce, Data 360, Marketing Cloud, SAP ECC, SAP Datasphere, HubSpot, Gmail and Google Calendar, Outlook, Google Drive, Google Sheets, SharePoint and OneDrive, Azure Storage, Google Cloud Storage, ServiceNow, Zendesk, Jira, Snowflake, BigQuery, Databricks, Microsoft Fabric and Azure Synapse, Trino and Starburst, Dynamics 365, Dynamics 365 Finance and Supply Chain, NetSuite, Oracle Fusion, Workday, SQL databases, MongoDB, Slack, Microsoft Teams, health records (FHIR)
Shared account Amazon S3, lake tables and Amazon Athena (your organization's AWS access); Gong (Gong gives the whole company's data); Gainsight and ChurnZero (one key or API user for the organization); custom connectors (one secret; your connector is told who is asking); SAP's public demo sandboxes for S/4HANA and SuccessFactors
The admin chooses per source Amazon Redshift (shared only when the admin names one shared database user); REST and SOAP sources (personal with OAuth, shared with an API key, token or password)

Add a shared source only where everyone who may use it may see what its account sees, and narrow who can use it with groups. The full list, with what decides each one, is in How people sign in to each system.

Reads, changes and query checks

  • Every query is checked before it is sent: one read-only statement in the system's own language, with anything that writes, deletes, locks, calls out or runs code refused. Databases are also queried inside read-only (or always rolled-back) transactions with a time limit. See How queries stay read-only.
  • Changes are off for every source until an admin turns them on for named objects. Then propose_change shows the current and new values, and nothing is sent until the person confirms: in SourceLace's own confirmation prompt where their app can show one, or otherwise by opening the confirmation link the app gives them, signed in to SourceLace, and typing the code it shows into their app. The AI app never sees that code, so it cannot confirm a change by itself. Three wrong codes discard the preview, and previews expire after 10 minutes. Where the system supports it, the change is refused if the record changed after the preview.
  • SourceLace never sends email: mail sources only make drafts, which the person sends themselves.

The audit trail

Every tool call is recorded, whether it succeeded, was refused or failed, from the SourceLace app, from every MCP client and from admins. So are sign-ins, every change to people, sources, groups and single sign-on, and every AI agent run, step and approval. A search across systems (Searching everything) writes one entry for each system searched, never the results. Failed sign-in and token attempts are recorded, including ones that cannot be tied to an organization, without the address or token itself; those are kept by SourceLace and are not part of your audit trail.

Each entry holds: when, who, the tool, the source, the person's id in that system, the query text, the object, the record id, the names of changed fields, how many rows came back, the outcome, the error type, and how long it took. It never holds row values, field values, passwords, keys or tokens. Note that the query text is kept as written, so a filter value typed into a query (such as a customer name) appears in the trail your admins read. It is stored encrypted with your organization's own key, and the Audit log's text search matches the person, action, source, object, record id and error rather than the words inside a query. The one exception is health record (FHIR) sources: their search parameter values often name a patient, so the trail keeps only the parameter names.

Entries are chained by SHA-256 hashes per organization: editing, removing or reordering an entry breaks the chain, which SourceLace detects. SourceLace checks the audit trail automatically every day and alerts your admins if anything does not match. You can also check it yourself at any time and export it with its proof. If an entry cannot be written, the call fails rather than returning data that was not recorded. Admins read the trail on the Audit log page. It is kept for your plan's audit history (90 days to 7 years); older entries are then deleted automatically, from the oldest, and SourceLace keeps the last deleted entry's hash so the entries kept still verify.

AI and your data

From your own AI app (MCP clients such as Claude, ChatGPT, Cursor or VS Code): SourceLace sends nothing to any AI model. It returns results to your AI app, which sends them to its own AI provider under your agreement with that provider.

From the assistant in the SourceLace app: to answer, SourceLace sends the conversation to Anthropic's API: the person's questions, the assistant's earlier replies, your project instructions and skills used, and the results of the tools the assistant called, which can include rows from your systems. Your admin chooses whether this goes through SourceLace's Anthropic account (the default), your organization's own Anthropic key, your organization's own OpenAI key, or your organization's own Google Gemini or Mistral key. With your own key it is under your own agreement with that provider. Requests to OpenAI are sent with OpenAI's store option off, so OpenAI does not keep the conversation for later retrieval.

From AI agents: an agent works like the assistant without a person typing. On its schedule, or when something happens in a connected system, SourceLace sends the agent's instructions and the results of the tools it calls to the same AI provider, acting as the agent's owner with the owner's own sign-ins and access. See AI agents.

SourceLace does not train AI models on your data, and does not send it to any AI provider other than the one your admin chose.

Data protection rules: your admin can hide fields and mask values such as email addresses, phone numbers and card numbers before any of these reach an AI, while the person who asked still sees the real values in the SourceLace app. See Data protection.

Hosting and subprocessors

  • Hosting: SourceLace's cloud is hosted in the United States: the application, its database and its backups, with backup copies in a second US region. The hosting provider is named in our privacy policy. How the hosting environment is protected, and how it is backed up, is in Security questions.
  • Subprocessors:
Company What for What it receives
Our cloud hosting provider (United States; named in our privacy policy) Hosting the application, its database, the short-term store for query results, file storage and backups; also sending SourceLace's own alert and notice emails Everything SourceLace stores (encrypted as described above) and processes, and the address and text of each alert or notice email it sends.
Anthropic The assistant in the SourceLace app (only when people use it) The conversation and tool results, as described in AI and your data.
OpenAI (only if your admin enters your own OpenAI key) The assistant in the SourceLace app, under your own agreement with OpenAI The conversation and tool results, as described in AI and your data.
Google Gemini or Mistral (only if your admin enters your own key from that provider) The assistant in the SourceLace app, under your own agreement with that provider The conversation and tool results, as described in AI and your data.
Microsoft Azure OpenAI or Amazon Bedrock (only if your admin connects your own resource or account) The assistant in the SourceLace app, in your own Azure resource or AWS account and region, under your own agreement with Microsoft or AWS The conversation and tool results, as described in AI and your data.
Google, Microsoft, or your single sign-on provider Signing people in to SourceLace The sign-in itself, under your own account with them.
Stripe Billing: card payments and invoices Your organization's name and billing contact's email address, and what is billed. Never data from your systems.
Our inbound email provider (named in the security review pack; only if you give an agent an email address) Receiving email sent to your agents' email addresses and passing it to SourceLace The emails your systems and partners send to those addresses, kept by the provider for a short time after they arrive (up to about two days), then removed under the retention SourceLace sets up with it.
Google Workspace SourceLace's company email Messages you send us, such as support requests. Never data from your systems unless you put it in an email.
Cloudflare Domain name service (DNS) for SourceLace's websites Only the lookups of our domain names; no data from your systems.

Your connected systems (Salesforce, SAP and so on) receive only the calls each person makes with their own login.

Certifications and agreements

SourceLace does not yet hold any security certification or attestation. SOC 2, ISO 27001 and HIPAA are not yet available, so neither is a HIPAA Business Associate Agreement. Hospitals and other HIPAA covered entities must not connect production patient data to a health record source until a business associate agreement is signed. A GDPR data processing agreement (DPA) is not yet available either. Data is hosted in the United States only today. A SOC 2 Type 2 examination is in progress; where that stands, and what is not in place yet, is in Organization and certifications.

For a security questionnaire, a review call, or to report a vulnerability, email security@sourcelace.com. What you can request is listed in What can we request for a security review?

Removing access and data

  • A person can disconnect any source at any time, which deletes their stored sign-in for it, and delete their own chats.
  • An admin can remove a person, remove a source, or turn changes off at any time.
  • Closing your account. To close your organization's account and have its data deleted, contact SourceLace at support@sourcelace.com. When SourceLace deletes your organization, nobody can sign in from that moment, and 7 days later everything SourceLace stores for it is removed for good, with a few exceptions: people, sources and sign-ins, chats, agents, plan and billing records, the audit trail and your organization's encryption key. During those 7 days the delete can still be undone, for example if it was a mistake. Some records follow their own schedules instead: the locked daily audit records are kept for 400 days, the records of requested web addresses for 365 days, and backup copies expire on the hosting provider's backup schedule (see the privacy policy).