One platform for asking, acting and governing.
SourceLace gives people and AI agents live, governed access to your business systems. Use it from your own AI app, from the SourceLace app, or through agents, and run it in our cloud, on your own servers, or both.
How SourceLace fits together.
People, AI apps and other systems ask. The platform answers and acts. Every request passes through the Trust Layer on its way to your systems, and changes come back to a person to approve.
Diagram: how SourceLace fits together Four layers from top to bottom: who asks, the SourceLace platform, the SourceLace Trust Layer, and your systems. Requests go down through the Trust Layer to your systems; answers come back up, and changes are made only after a person approves them. The same content follows as a list.- Webhooks from systems such as Salesforce, SAP Event Mesh or MuleSoft Preview
- Agents on platforms that support the open A2A standard coming
- On a schedule, before meetings, or when something happens
- Shadow mode and approvals
- Team agents that run for each person as themselves
- One question, one exact answer from several systems
- Each system asked only for what it holds
- Where each number came from
- Setup assistant
- Industry packs
- Readiness score
- Each person's own sign-inTheir own permissions in each system
- Access by groupAdmins choose who may use which systems
- Data protectionMasked values; real ones only for people you allow
- Changes with a person's yesEvery change previewed; a person approves unless an admin allows that action
- Tamper-evident audit trailWho did what and when, chained
- Nothing copiedResults kept briefly, never stored as a copy
- Salesforce
- HubSpot
- Dynamics 365
- ServiceNow
- Zendesk
- Jira
- SAP S/4HANA
- SAP ECC
- SuccessFactors
- Oracle Fusion
- NetSuite
- Workday
- Snowflake
- BigQuery
- Databricks
- Redshift
- Fabric
- Google Drive
- Gmail
- Microsoft 365
- SharePoint
- Slack
- Teams
- Any REST or SOAP API
- Connector SDK
Runs where it suits you: in our cloud, on your own servers, or bothRead live, where the data already is
Who asks
- People in the SourceLace app: chat, projects, charts and reports at sourcelace.ai, with each person's own access.
- Your own AI apps: any AI app that supports the Model Context Protocol (MCP), added with one address.
- Other systems and agents: webhooks from systems such as Salesforce, SAP Event Mesh or MuleSoft (Preview), and agents on platforms that support the open A2A standard (coming).
Questions and requests go down to the platform; answers, drafts and proposed changes come back up.
SourceLace platform
- AI agents (Preview): on a schedule, before meetings, or when something happens; shadow mode and approvals; team agents that run for each person as themselves.
- Exact answers across systems (Preview): one question, one exact answer from several systems, each system asked only for what it holds, with where each number came from.
- Setup that does itself (Preview): setup assistant, industry packs and readiness score.
SourceLace Trust Layer, which every request passes through
- Each person's own sign-in, with their own permissions in each system.
- Access by group: admins choose who may use which systems.
- Data protection: masked values, with real values only for people you allow.
- Changes with a person's yes: every change previewed and approved first.
- Tamper-evident audit trail: who did what and when, chained.
- Nothing copied: results kept briefly, never stored as a copy.
Reads go live to your systems as each person; changes are made only after a person approves.
Your systems
- CRM and service: Salesforce, HubSpot, Dynamics 365, ServiceNow, Zendesk, Jira.
- ERP and HR: SAP S/4HANA, SAP ECC, SuccessFactors, Oracle Fusion, NetSuite, Workday.
- Data platforms: Snowflake, BigQuery, Databricks, Redshift, Fabric.
- Files, mail and chat: Google Drive, Gmail, Microsoft 365, SharePoint, Slack, Teams.
- Your own systems: any REST or SOAP API, and the Connector SDK.
SourceLace runs in our cloud, on your own servers, or both.
Bring the AI app your team already uses.
SourceLace is an MCP connector that any AI app supporting the Model Context Protocol can add with one address. It works with any AI model: SourceLace never assumes a particular vendor, and when people use their own AI app, SourceLace sends nothing to an AI model itself.
The same tools for every system
Find the right objects and fields, including your custom ones, query them live in each system's own language, read a record with its related lists, and propose changes.
Signed in as themselves
People sign in to SourceLace once in their AI app, then connect each system with their own account wherever the system offers one.
Your how-tos travel with you
Skills your organization writes, such as "build a QBR deck", are available in every AI app, in the SourceLace app and to agents.
Answers about SourceLace itself
The AI can look up setup, security and privacy questions in SourceLace's own manual and answer with a link to the page.
A workspace for your data questions, at sourcelace.ai.
People sign in, connect their systems and chat with an assistant that uses the same tools, with their own permissions. It uses the AI included in your plan, or your organization's own AI key (Anthropic and OpenAI keys today).
Chats and projects
Group chats into projects and share them. Chats are saved, encrypted, and deleted after your organization's retention period.
Charts
Turn a result into a chart. Pinned charts refresh themselves.
Files
Make Word, PowerPoint and PDF reports from results, ready to share.
Skills and voice
The assistant follows your team's written skills. Ask by speaking and have answers read aloud.
One exact answer, even when it spans systems.
Ask a question that needs several systems at once, such as Salesforce accounts with overdue SAP invoices, and get one exact answer. Each system is asked only for what it holds, every row is counted, and you can see where each number came from. Follow-up questions can work on the same result without fetching everything again.
- "Which Corvanta renewals this quarter have overdue invoices?"CRM + ERP · one answer across both
- "Of those, which have an open severity 1 case?"adds the service desk · works on the result you already have
Changes with a preview, and only after a yes.
Reading is read-only and checked before anything is sent. Where an admin allows changes for a system and object, the AI proposes each change, the person sees exactly what would change, and it is applied once, only after they confirm.
- Off until an admin turns it on. Changes are allowed per system and object, and only for the groups an admin chooses.
- Previewed field by field. Every field's current and new value, for records, spreadsheet rows, and Slack or Teams posts alike.
- Drafts, never sent mail. Mail connectors create drafts for the person to review and send.
- Recorded. Who changed which fields, on which system, and when.
Everything your admins need, in the app.
Sign-in and single sign-on
Google or Microsoft accounts out of the box, or your own identity provider with OpenID Connect or SAML 2.0 SAML: Preview. Require single sign-on for your domains and sync groups.
Automatic provisioning Preview
With SCIM, Microsoft Entra ID, Okta or Ping Identity add people, keep their groups up to date, and deactivate them when they leave. Their sessions, AI app connections and sign-ins end at once.
People, groups and sources
Invite people or allow your email domains, decide which groups may use which systems and objects, and add or change sources.
Limits and usage
See every limit on one page and make some stricter, such as shorter sign-in sessions or fewer agent runs at once. Usage and invoices are in the app too.
Retention and audit
Shorten how long results, chats and agent runs are kept. Search and filter the audit trail.
Data protection
Hide fields from the AI or mask sensitive values, per system and object. How it works for you.
A good setup, suggested for you and approved by you.
SourceLace helps admins get to a sound setup quickly. Every suggestion waits in a review queue with its reason, and nothing changes until an admin approves it.
Readiness score
A score from 0 to 100 with the checks behind it: sign-in, access, data protection, sources, AI model, agents, skills and terms, and audit and retention. Most checks offer a fix.
Setup assistant
After you connect a system, SourceLace suggests groups, data protection rules for fields that look sensitive, business terms such as ARR, and agents that fit.
Industry packs
Ready-made starting points for RevOps on Salesforce, supply chain on SAP, HR on SuccessFactors or Workday, and healthcare on FHIR with test data.
Business terms
Tell SourceLace what your words mean, such as what counts as open pipeline, so every answer uses your definitions.
Setup as a file
Download your whole setup as one file to keep a record, review it, or copy it from a test organization to your real one. It never holds people, passwords, keys or data.
Learn from use
Off until an admin turns it on. SourceLace notices questions many people keep asking and suggests a skill or an agent for them, without showing who asked.
Run it where it suits you.
SourceLace cloud
Sign in at sourcelace.ai and start. Each customer is its own organization, with its own encryption key for what SourceLace keeps.
SourceLace Runner
Stay in our cloud, and run the AI agents you choose inside your own cloud account or data center. The runner connects out over HTTPS only, so you open no inbound port, and run results stay in your network. Turned on per organization, on request.
Your own servers
Run SourceLace yourself with Docker or Kubernetes, from the same software as our cloud, with a license file from SourceLace. You hold the keys.