Privacy Policy
Effective 18 November 2026
This policy explains what information SourceLace collects, why, where it is kept and for how long. SourceLace is operated by SourceLace, Inc., a Delaware corporation, 1734 Holmes Dr., Sewickley, PA 15143, USA ("SourceLace", "we", "us"). It covers the website sourcelace.com, the SourceLace app at sourcelace.ai, and the SourceLace MCP connector that people add to AI apps such as Claude or ChatGPT (together, the "Service").
The short version
- SourceLace reads your business systems live and does not copy their data into a store of its own. The one exception is your organization's choice: if it adds Saved answers and your admin turns them on for a system, each person's own recent answers from that system are kept in memory, encrypted with a key unique to your organization, for at most the time your admin sets (up to 24 hours), and are never shown to anyone else. Without that choice, every answer is read live. For most systems each person signs in with their own account, so the system's own permissions apply.
- Query results are held for up to 30 minutes so the AI can use them, in server memory or, encrypted with a key unique to your organization, in a short-term store or, for large answers, an encrypted file in short-term storage, then discarded. The exceptions are chats in the SourceLace app and AI agents, described next.
- Chats in the SourceLace app are saved automatically, including the answers and the rows they show, so you can come back to them. They and the files made in them are encrypted with a key unique to your organization and deleted after your organization's retention period (30 days after a chat was last used, unless your admin sets another).
- AI agents your organization sets up keep a summary of each run and the changes waiting for someone's approval, which include the record's current and proposed values. These are encrypted the same way and deleted with the agent or after the same retention period.
- We never sell personal information and never use your data, or data from Google, to train AI models or for advertising.
Who this applies to
Most people use SourceLace through their employer or another organization that has an account (a "Customer"). For data that flows through the Service from a Customer's business systems, the Customer decides what is connected and who may use it, and we process that data on the Customer's behalf and on its instructions. Questions about that data should go to the Customer's administrator first. For our own website and for account and billing records, we decide how the information is used.
What we collect
- Account information. Your email address, the organization you belong to, your role there, who invited you, and when you last signed in. You sign in with Google, Microsoft or your organization's single sign-on; SourceLace asks them only for your name and email address and keeps the email address. If your organization provisions people through SCIM, SourceLace also keeps the profile your identity provider sends (such as name, title, department and manager), encrypted, while you are a member.
- Connection tokens and logins. When you connect a business system, that system gives SourceLace a sign-in token for you. For most systems you sign in on the system's own page, so we never see your password. For databases, which have no such page, you type your database login into SourceLace. Tokens and logins are encrypted with a key unique to your organization, used only to act for you, never shown back and never written to logs. A few systems an admin may add, such as Gong, Amazon S3 or a connector your organization built, use one shared access set up by the admin instead of a sign-in per person.
- Business data you ask about. Records, files, emails and other content returned by your connected systems when you, or an AI app acting for you, asks a question. It is read on demand and held for up to 30 minutes, in server memory or, encrypted with a key unique to your organization, in a short-term store or, for large answers, an encrypted file in short-term storage. Files are read where they sit and never indexed. A change you are asked to confirm is held the same way for 10 minutes, then discarded if not confirmed. Saved answers (only if your organization chooses them): in memory only, for the person who asked, for at most the time your admin sets, up to 24 hours. When you ask a question that combines systems you have connected, SourceLace may send values from one of them (for example, the values of a column used to match records) to another as a filter, so the second system returns only the rows needed. Values are sent only to systems you have connected, and SourceLace does not keep them.
- Chats and what you make in the app. Chats in the SourceLace app, saved automatically (your questions, the answers and the data shown in them), files generated in a chat (such as Word, PowerPoint or PDF reports), charts pinned to a project, projects with their instructions and the people they are shared with, and skills your organization writes.
- Photos in chats (only if your organization's admin turns them on). If your organization's admin turns on photos, a photo or screenshot you add to a question is sent to the AI provider (see AI processing) so it can answer, and is kept encrypted with your chat until your organization's chat retention period ends. Photos and screenshots cannot be checked by your organization's data protection rules, which is why photos are off unless an admin accepts responsibility for them, and your admin chooses who may use them. Deleting a chat deletes its photos from SourceLace's storage; an older stored copy can remain for up to 30 days, and backups follow their own schedule, and the copy the AI provider holds follows that provider's own terms. Each acceptance is recorded in your organization's audit log.
- AI agents. Agents your organization sets up: their instructions and settings, a record of each run (times, which tools and systems were used and how many rows came back, never the rows themselves), an encrypted summary of each run and what triggered it, and changes waiting for approval, which include the record's current and proposed values. Pending approvals expire after 7 days unless your admin sets another period, never more than 30; runs and approvals are deleted with the agent or after your organization's retention period.
- Agent Architect (only if your organization uses it). Your answers about your job (such as your role, your duties and what slows you down), a record of each run (when, which sources, the data level and its AI cost), and the agent proposals written for you, which can name objects and fields in your systems but never contain values from your records. Answers and proposals are encrypted with a key unique to your organization. When an admin accepts Agent Architect's statement, the name they type is kept with the acceptance and in your organization's audit records. If you link files, as examples of your work or as references for an agent, we keep a link to each linked file with its name and outline (such as its headings), encrypted with your organization's key, never a copy of the file; the text read from a file for a run is deleted when that run ends. Files you upload are described under Uploaded files.
- Uploaded files (only if your organization allows uploads). Files your users upload as references for their agents or as examples of their work in Agent Architect, and the pictures they share in Agent Architect where your administrator allows pictures. We keep them encrypted with a key unique to your organization and use them only to design and run your agents.
- We also keep each person's own choices about how the app looks and what it shows them, such as which projects they have pinned. These hold no content from your chats or your systems, and are deleted when the person is removed or their organization closes.
- Business and billing contacts. For organizations that are customers or are considering SourceLace: a contact's name, email address and phone number, notes from our conversations, and the email addresses an admin enters for health alerts. We keep them while the organization is a customer, and delete them 12 months after a prospect is marked lost (this covers the contact's details, our notes, discovery answers and health alert email addresses; it does not change the sign-in accounts of the organization's people, which are removed when the organization is deleted, or email addresses that appear in its audit log, which follow the audit log's retention period). We remove everything when the organization is deleted.
- Organization settings. Settings your admins choose, such as chat retention and which AI model to use, and, if your organization brings its own AI provider key or single sign-on, that key or secret, stored encrypted and never shown again. Admins also describe the interfaces of their own systems, so that people can ask about them: the names of an API's operations, parameters and fields, and which parameters narrow which fields. These are descriptions, not data from the systems, and they are stored with the connection.
- Audit records. For every request: who made it, when, which tool and system, the object and record, the query text, the names of fields changed, how many rows came back, and whether it succeeded, including the runs of AI agents and who approved their changes. Changes SourceLace's own staff make to your organization are recorded the same way, under their name. The query text can include values you typed, such as a customer name; audit records are stored encrypted with your organization's own key. An Audit log export, and a SourceLace staff member reading your audit log, are recorded too, with the filters used, including any search word. Audit records never contain the rows returned or the values written. Once a day we also save a short record for each organization (its identifier, the number and fingerprint of its latest checked audit entry, the time and whether the check passed) in storage that cannot be changed or deleted for 400 days. It holds nothing about people or about your systems.
- Usage and billing information. Counts of requests per person and the AI usage of each question (token counts only, never the question or the answer), used to apply plan limits and to bill. Activity counts per organization and day (how many people were active, questions asked, agent runs and the like), used to run and improve the Service: counts only, never your data, kept in SourceLace's own database and never sent to an outside analytics or tracking service. Payments are handled by Stripe, which receives your organization's billing contact and company name, never data from your systems; we do not store card numbers.
- Website and cookies. sourcelace.com sets no cookies and has no analytics. Its pages load fonts from Google Fonts, so your browser connects to Google's servers. The app at sourcelace.ai uses one cookie to keep you signed in. If you email us, for example to request a pilot, we keep that correspondence. Our servers and firewall record IP addresses and browser details in security logs, which we keep for one year.
How we use it
- To provide the Service: sign you in, connect your systems, answer your questions, make the changes you confirm, and show your organization's admins its audit trail and usage.
- To keep the Service secure, prevent abuse and investigate problems.
- To bill your organization and enforce its plan.
- To contact you about the Service, such as security notices and changes to these terms.
We do not sell personal information, do not use it for advertising, and do not use your data or your systems' data to train AI models.
AI processing
In this policy, "the AI provider" means the provider that writes the answers for your organization: Anthropic, through SourceLace's own account by default, or the provider whose API key or cloud account your organization's admin connected, as described below.
If you use SourceLace from your own AI app (for example Claude or ChatGPT), SourceLace returns results to that app and sends nothing to an AI model itself; your app's own provider and terms apply. The assistant inside the SourceLace app sends your question, the chat so far and the relevant results to Anthropic, which provides the AI model that writes the answer: either through SourceLace's own Anthropic account, or through your organization's own Anthropic API key if your admin entered one, in which case your organization's agreement with Anthropic applies. Under Anthropic's commercial terms, Anthropic does not train its models on data sent through its API. If your admin chooses your organization's own OpenAI API key instead, the same content goes to OpenAI under your organization's agreement with OpenAI, and SourceLace asks OpenAI not to store it for later retrieval (OpenAI's own policies may still keep it for a limited time, for example for abuse monitoring). If your admin chooses your organization's own Google Gemini or Mistral API key, the same content goes to that provider under your organization's agreement with it, and that provider's own terms decide how long it is kept. If your admin connects your organization's own Azure OpenAI resource or Amazon Bedrock account instead, the same content goes to that service in your organization's own Azure resource or AWS account and region, under your organization's agreement with Microsoft or Amazon Web Services.
To answer faster and at lower cost, the AI provider may keep a processed copy of the conversation in its memory for a short time and then delete it. For example, with Anthropic today this is about 5 minutes after your last question, and up to an hour for the fixed instructions that start every chat (SourceLace's and your organization's), and Anthropic states that these caches are never shared between organizations. Other AI providers cache under their own terms, and we update the list of providers on this page before adding one that handles your data.
AI agents work the same way without a person typing: on their schedule, or when something happens in a connected system, SourceLace sends the agent's instructions and the data it reads to the same AI provider, acting as the agent's owner.
Your organization's admins can set data protection rules that hide chosen fields from the AI and replace values such as email addresses, phone numbers and card numbers with placeholders before anything is sent to an AI model, whether from your own AI app, the SourceLace app or an AI agent. What each placeholder stands for is held like a query result, for at most your organization's result time limit, and is never written to the audit trail.
When your organization uses Agent Architect, SourceLace reads the structure of the data sources you choose, works out summary figures, and, if your administrator allows it, reads a small sample of rows, and sends these to your organization's AI provider so it can write agent proposals. If a person shares examples of their work, or adds files to an agent as references, by linking them or uploading them, SourceLace reads those files (linked files with that person's own access) and sends their text and layout to your organization's AI provider, when Agent Architect designs agents and each time such an agent runs. Anything your organization's data protection rules, or any extra masking rules your administrator sets for Agent Architect, mask is replaced by a placeholder before anything is sent, and the AI provider never receives the original value. Pictures cannot be masked: if your administrator allows pictures, a picture a person shares is sent to your organization's AI provider as it is. Data protection covers what your organization has set up, so your administrator should review those rules before allowing sample rows or shared files. Samples, summary figures and the text read from shared files are held briefly and then deleted, and saved proposals contain no sample values. A linked file is kept only as a link with its name and outline. An uploaded file is kept, encrypted with a key unique to your organization, until the person removes it or the agent or conversation it belongs to is deleted.
If your organization runs some AI agents on a SourceLace Runner inside its own network, those agents' AI calls use your organization's own AI key from that network, and their summaries, drafts and changes waiting for approval stay on the runner. SourceLace receives only each run's status, times, the tools and systems used and counts.
If you use voice input in the app, your browser turns your speech into text with its own speech service (in Chrome and Edge that is Google's or Microsoft's; Safari uses Apple's or the device). SourceLace receives only the text.
Data from Google
If you sign in with Google, SourceLace asks only for your name and email address. If you connect Google sources, SourceLace requests only the access each source needs: reading mail and creating drafts in Gmail (it never sends email), reading Google Calendar and, only where your organization's admin turns it on and only after you confirm each one, making, changing and cancelling events you organize, reading Google Drive files (and, only where your organization's admin lets agents save the files they make, adding new files to the folder they name), reading and changing the Google Sheets you name, and running read-only BigQuery queries. SourceLace's use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements. In particular:
- We use Google data only to provide the features you use in SourceLace, at your request.
- We do not transfer Google data to others except as needed to provide those features (for example to the AI model provider that writes the answer you asked for in the SourceLace app, or to the AI app you use SourceLace from), to comply with law, or as part of a merger or acquisition with notice to you.
- We do not use Google data for advertising, and we do not sell it.
- No person at SourceLace reads your Google data unless you ask us to for support, it is needed for security or to comply with law, or it has been aggregated and anonymized for internal operations.
- We do not use Google data to develop, improve or train generalized AI or machine learning models.
Google data is handled like any other source data: held for up to 30 minutes, and kept longer only inside chats in the SourceLace app and AI agents' run summaries and pending changes, as described above. You can remove SourceLace's access at any time by disconnecting the source in SourceLace, or from your Google Account's third-party connections.
Data from Slack
If you connect Slack, you sign in with your own Slack account and SourceLace acts as you. There is no SourceLace bot in your workspace and no shared account, so SourceLace sees only what you can see in Slack. It asks Slack only for the access it uses: listing the channels, direct messages and group messages you are in; reading their messages, threads and the files shared in them; searching messages; and looking up people's names and email addresses so answers show names and can tell who is outside your organization. Where your organization's admin allows it, SourceLace can also act for you: posting messages, replies and reactions, creating channels and canvases, and uploading files, such as a report an AI agent made. In a chat you confirm a preview before anything is posted; an AI agent posts only where it was set up to deliver.
- We use Slack data only to provide the features you use in SourceLace, at your request.
- We do not transfer Slack data to others except as needed to provide those features (for example to the AI model provider that writes the answer you asked for in the SourceLace app, or to the AI app you use SourceLace from), to comply with law, or as part of a merger or acquisition with notice to you.
- We do not use Slack data for advertising, and we do not sell it.
- We do not use Slack data to develop, improve or train AI or machine learning models.
- No person at SourceLace reads your Slack data unless you ask us to for support, or it is needed for security or to comply with law.
Slack data is handled like any other source data: held for up to 30 minutes, and kept longer only inside chats in the SourceLace app and AI agents' run summaries and pending changes, as described above. Your Slack sign-in token is encrypted with a key unique to your organization and never shown back. You can remove SourceLace's access at any time by disconnecting Slack in SourceLace, and a Slack workspace admin can remove the SourceLace app from the workspace.
Where it is stored and who helps us
The Service is hosted in the United States. We use these service providers, each bound by contract to protect the data and use it only to serve us:
| Provider | What for |
|---|---|
| Amazon Web Services | Hosting the Service, the website, the database, the short-term store for query results, storage for large results and archived records, and backups; also sending SourceLace's own emails (alerts and notices) (United States) |
| Anthropic | AI model for the SourceLace app's assistant, when your organization uses the AI included in its plan |
| OpenAI | AI model for the SourceLace app's assistant, only when your organization's admin enters the organization's own OpenAI key |
| Google (Gemini), Mistral | AI model for the SourceLace app's assistant, only when your organization's admin enters the organization's own key from that provider |
| Microsoft (Azure OpenAI), Amazon Web Services (Bedrock) | AI model for the SourceLace app's assistant, only when your organization's admin connects the organization's own Azure OpenAI resource or Amazon Bedrock account; the content then goes to that service in your organization's own account and region |
| Google Workspace | Our company email, including messages you send us |
| Cloudflare | Domain name service (DNS) for our websites |
| Stripe | Billing, card payments and invoices (receives your organization's billing contact and company name, never data from your systems) |
The business systems you connect, and the AI app you use SourceLace from, are chosen by you or your organization and are governed by your own agreements with them. We will update this list before adding a provider that handles Customer data.
How long we keep it
- Query results: up to 30 minutes. Unconfirmed change previews: 10 minutes. Both are held in server memory or, encrypted, in a short-term store or, for large answers, an encrypted file in short-term storage, and are written to the database only as part of a saved chat.
- Saved answers (only if your organization chooses them): in memory only, for the person who asked, for at most the time your admin sets, up to 24 hours. They are deleted sooner when the admin turns them off for a system or the person changes something in it through SourceLace.
- Chats and the files made in them: until you delete them, or automatically after your organization's retention period, counted from the chat's last use (30 days unless your admin sets another). Charts pinned from a chat go with it.
- AI agent runs and their pending changes: deleted with the agent, or after your organization's retention period. A change still waiting for approval expires after 7 days unless your admin sets another period, never more than 30.
- Agent Architect: your answers, runs and proposals until you delete the run, you leave your organization, or the organization closes. After that, only a record of the run's time and AI cost stays, with no person or content, so your organization's allowances still count it. Samples and summary figures are deleted when the run ends. An admin's acceptance is kept until it changes or the organization closes, and in the audit records for your plan's audit history. A link to a referenced file, with its name and outline, is kept until it is removed from the agent or the agent is deleted; a run's list of the files it used is kept with that run.
- Uploaded files: An uploaded file is kept until the person removes it, or until the agent or the Agent Architect conversation it belongs to is deleted; the files of a conversation that is never used are deleted after 14 days. Deleted files are removed from our storage promptly, and from our backups when those expire; every file is deleted when your organization leaves SourceLace.
- Projects and skills: until you or your admin delete them.
- Connection tokens and logins: until you disconnect the system, the token expires, or your account is removed.
- Organization settings and the descriptions of your systems: until your admin changes or removes them, or the organization closes.
- Records of requested web addresses. The servers in front of SourceLace record the web address each request asked for, and keep those records for 365 days. Words you type into the search box of some lists in the app can be part of that address.
- Audit records and usage counts: for your plan's audit history (90 days to 7 years), because the audit trail is your organization's tamper-evident record.
- The daily audit records described above: 400 days, locked.
- When an organization closes its account, we delete its data 7 days later (the 7 days allow the closing to be undone), except where law requires us to keep something longer. Copies in the hosting provider's backups expire on its backup schedule.
Security
Data is encrypted in transit (HTTPS) and at rest. Connection tokens and logins, chats, generated files, projects, skills, AI agents' run summaries and pending changes, saved answers, and organizations' own AI keys are additionally encrypted with a key unique to each organization, which SourceLace manages. Every request is recorded in a hash-chained audit trail, so a changed or deleted entry shows. Where you sign in to a system with your own account, you see only what your permissions there allow; for shared-access sources an admin adds, admins can limit who may use them. No system is perfectly secure; if a breach affects your information, we will notify you and your organization as the law requires. SourceLace does not hold any security certification such as SOC 2 or ISO 27001.
Your choices and rights
You can disconnect any system, delete your chats, and ask us for a copy of your personal information, to correct it, or to delete it. Depending on where you live (for example the EU, UK or California), you may have further rights, including to object to or restrict processing and to complain to your data protection authority. If your account belongs to a Customer, we may refer your request to that Customer. Email privacy@sourcelace.com; we answer within 30 days.
If you are in the EU or UK, we rely on contract (to provide the Service), legitimate interests (security and improving the Service) and legal obligation. Where information moves outside your country, we use safeguards such as the European Commission's standard contractual clauses.
Children
The Service is for businesses and is not directed to anyone under 16. We do not knowingly collect their information.
Changes
We will post changes here with a new effective date, and tell account holders by email before a material change takes effect.
Contact
SourceLace, Inc., 1734 Holmes Dr., Sewickley, PA 15143, USA. Email privacy@sourcelace.com. Security reports: security@sourcelace.com.