Connect sources once
Website pages, uploaded files, and Google Drive build one workspace-scoped knowledge base today; Notion, Slack, Zendesk, Intercom, SharePoint, and OneDrive sync are on the way.
The permission-aware answer layer
Cited answers from your docs — with source permissions enforced before retrieval.
Braize answers questions from your website and your files with a citation on every claim — and enforces each source's permissions inside the search, before an answer is generated. Add a source once, then use it from the dashboard, a REST API, an OpenAI-compatible endpoint, and MCP (beta) for your agents.
Free while in beta — no credit card. Hosted in Europe.
Sources live now:WebsiteFilesDriveComing soon:NotionSlackZendeskIntercomSharePointOneDrive
Watch it work
How Braize works, in thirty seconds
Sources connect, you ask a question, and a cited answer arrives the same answer over the dashboard, the REST API, and MCP — no glue. The controls below let you scrub, pause, or restart.
The product, plainly
What is our product?
Braize is a permission-aware answer layer that turns your company sources into permission-aware, cited answers your developers and agents call over a REST API, an OpenAI-compatible endpoint, and MCP.
How does it help?
It removes the operational work of keeping your answers accurate, permission-safe, and traceable to the source of truth.
Keep knowledge current
Re-ingest a source and its answers update; every document carries a freshness score and a last-indexed status. (Automated change detection, conflict alerts, and staleness prediction are on the roadmap.)
Filter before retrieval
Workspace, region, user identity, groups, and public scope travel with every request, so unauthorized content is never retrieved.
Answer with proof
Every response is grounded in source documents and returned with citations that customers, employees, and auditors can inspect.
Serve your surfaces
The same brain answers over the REST API and the MCP agent endpoint today; the web widget, WhatsApp, a Slack app, and a helpdesk feed are on the way.
Stay region-bound
Each workspace has one isolated regional data store; processing policy controls where inference is allowed to run.
Three jobs, one product
What Braize does
Connect your knowledge
Live today: your Website, Files, and Google Drive — permission-aware. Coming soon: Notion, Slack, Zendesk, Intercom, SharePoint, OneDrive.
Answer where you build
Live today: a REST API and an OpenAI-compatible endpoint. MCP for agents is in beta; the web widget, WhatsApp and Slack apps are on the way.
Keep it current
Re-ingest a source and Braize updates its answers. Every document carries a 0–100 freshness score and shows its last-indexed status, so you can see what's current. (Automated change-detection is on the roadmap.)
How it works (the technical version)
Sources → Permission-aware retrieval → Cited answers
Each stage is independent, so we can scale or replace one without disturbing the others. Region and workspace context travel with every request — your data never lands in the wrong place.
Sources
Website crawler
Files (PDF/DOCX/…)
Google Drive
Notion
Slack (read)
Zendesk / Intercom
SharePoint / OneDrive
Pipeline
Parse + Chunk + Embed
vision · forms · tables · content-hash
Freshness Engine (roadmap)
diff-aware reindex · conflict triage
ACL pre-filter
identity · groups · public-scope
Brain (workspace)
Isolated data store
docs · embeddings · ACLs
Hybrid search
keyword · semantic · context
Answer + cite
region-bound · ZDR-aware
Groundedness check
every answer · 100%
Channels
Web widget
Slack app
REST API
Agent endpoint
Helpdesk feed
Infrastructure
SSO + identity
standards-compliant · per-workspace
Scheduled jobs (roadmap)
freshness · reindex · conflict triage
Event bus
transactional · idempotent · auto-retry
Outbound proxy
SSRF guard · safe fetches
Edge routing
regional · fail-over · cached
Connect your knowledge
Live today — more sources on the way
Website, Files, and Google Drive ingest now; Notion, Slack, Zendesk, Intercom, SharePoint and OneDrive are coming soon.
- LiveWebsiteSingle-URL ingest
- LiveFilesPDF · DOCX · XLSX · CSV · MD
- LiveGoogle DrivePermission-aware sync
- Coming soonNotionPages + databases
- Coming soonSlackChannels + threads
- Coming soonZendeskArticles + tickets
- Coming soonIntercomHelp center + chats
- Coming soonSharePointSites + document libraries
- Coming soonOneDriveFiles + folders
Answer surfaces
Live today — the REST API (MCP in beta)
The REST API serves answers now and MCP is in beta; the web widget, WhatsApp and Slack apps are on the way.
- LiveREST APItyped contracts · streaming · OpenAI-compatible
- BetaMCPagent-ready, ACL-safe retrieval
- Coming soonWeb widget<50KB · streaming
- Coming soonWhatsAppofficial messaging API
- Coming soonSlack@mention + DM
- Coming soonHelpdesk feedsearch integration
Who Braize is for
Four teams, four problems Braize solves
Support and operations teams
E-commerce and customer support
When a policy changes, the answers customers get should change with it — no stale quotes, no manual re-training.
When you re-ingest the changed document, Braize updates its answers and links every answer back to its source file. Cited answers serve over the REST API today; the web widget and WhatsApp are on the way.
REST API · MCP · widget (soon)
Platform and integration engineers
B2B SaaS and developer platforms
Expose your knowledge to customers' AI agents without building and maintaining several separate auth flows.
Braize exposes a secure, region-pinned API endpoint that AI agents can discover and use. Permissions travel with each request — agents see only what the end user is allowed to see, enforced per request.
REST API · MCP
Internal-tools and IT teams
HR, IT, and employee enablement
Employees ask public questions in chat, but some answers should only reach people allowed to see private folders.
Braize enforces access permissions at answer time as a pre-filter inside retrieval — each employee sees only what their identity is allowed to access, never more. Google Drive syncs those permissions today (per-file sharing grants); Notion and Slack sync are on the way.
REST API · MCP · Google Drive sync
Security and compliance leaders
Regulated and data-sensitive industries
Keep both data and inference inside an approved region, and restrict processing to zero-data-retention endpoints.
Each workspace is one isolated store in one region. Choose which regions inference may run in and require zero-data-retention endpoints — enforced as a hard rule, never a silent fallback.
API · MCP · region-pinned
Every feature, in one place
The full inventory
Pick a category to dive in. Every feature lists the tier it becomes available at.
Ingestion
8 featuresf1Website crawler
Sitemap discovery, JS-rendered pages, learned-cadence re-crawl.
LiveFree+f2File upload
PDF / DOCX / XLSX / CSV / MD / HTML / images, malware-scanned at ingest.
LiveFree+f3Google Drive sync
Permission-aware sync: mirrors each file's explicit Drive sharing grants into the ACL pre-filter, with incremental + real-time updates. Files shared only via domain-wide or "anyone with the link" access aren't indexed yet.
LiveStarter+f4Notion connector
Planned: page + database selector with workspace-visible ACL (disclosed).
Coming soonStarter+f5Slack as a source
Planned: channel selector that will honor membership ACLs and index threads.
Coming soonStarter+f6Zendesk / Intercom feed
Planned: help-center articles + opt-in tickets, with PII redaction before indexing.
Coming soonGrowth+f40SharePoint / OneDrive sync
Planned: Microsoft 365 document libraries, site pages, and personal drives with permission-aware sync (per-item ACLs, delta updates).
Coming soonStarter+f31Outbound security proxy
Blocks malicious outbound fetches at the network boundary, including DNS-based attacks.
LiveFree+
Honest pricing
Free while in beta
No paid plans yet, no credit card. Paid tiers are on the way — join the waitlist below.
Free · while in beta
$0no credit card
- Your Website, Files, and Google Drive as sources
- Cited, grounded answers via the dashboard, REST API, OpenAI-compatible endpoint, and an MCP server (beta)
- Permission-aware retrieval — access enforced before the search
- Multilingual answers with citations on every claim
- Fair-use limits while in beta
Paid plans · coming soon
Higher limits, more channels, and connectors are on the way. Join the waitlist and we'll tell you the moment paid plans open — and ask what you need first.
Built for sovereign teams
UAE PDPLEU GDPREU AI Act Art 50ZDR-aware routingRegion-pinned data
Hosted in Europe. Each workspace is pinned to one region; embeddings never leave it.
FAQ
How does Braize avoid making things up?
Every factual claim links back to its source. A groundedness check audits every answer; ≥90% grounded is the target. When the knowledge base doesn't cover a question, Braize says 'I don't know' rather than guess.
What does 'region-pinned' actually mean?
Each workspace lives in one isolated data store in your chosen region. Embeddings — which are personal data — never leave that store. Logs, analytics, backups, none of it.
How fresh is the answer?
Re-ingest a source to update its answers — the answer cache clears when the document changes, not on a clock. Every document carries a 0–100 freshness score and shows its last-indexed status. Automated change-detection across your sources is on the roadmap.
Can Braize see my employee's private files?
Only what they're allowed to see. Braize checks permissions before it searches — the access filter lives inside the retrieval query, not after the fact — so people only ever see content their assigned access already lets them see. As source connectors land, they mirror the source's own permissions into that same pre-filter.
Which surfaces can I use today?
Configure the knowledge once, then use it through the dashboard, a REST endpoint, an OpenAI-compatible endpoint, or the agent endpoint over MCP (in beta). The web widget, WhatsApp, and Slack apps follow as they ship. Same brain, every mouth.
What about hallucinations from prompt injection?
Ingested content is scrubbed before indexing — HTML stripped, hidden instructions removed. Retrieved content is then treated as data, not instructions, at answer time.
How do I switch language?
Open Profile, then Language. You can choose English (the default) or Persian (Farsi, with right-to-left text). Your choice follows you across devices and is saved to your profile.
Why one region per workspace?
Data residency is built in. When you pick a region for your workspace, we enforce that choice — your data and AI inference stay in that region, and we refuse unsafe alternatives rather than falling back silently. This keeps you clearly aligned with UAE PDPL and EU GDPR.