Knowledge Bases

Your private,
queryable second brain.

Point Definable at your documents, databases, websites, repos, or ticketing systems. We turn them into a searchable, grounded knowledge layer — with citations, live sync, and tenant-level isolation.

Create a knowledge base How ingestion works
Any sourceLive syncAlways cited
Brains · 12
PDF
Policies 48
DB
Customer DB 12k
WEB
Support KB 248
GIT
Monorepo 12k
DRV
Brand assets 324
API
Shopify orders live

Customer DB

postgres://prod · 12,418 rows · synced 2m ago
Which of our enterprise customers haven't been contacted in 45+ days?
8 enterprise customers haven't been contacted in the last 45 days. Highest ACV: Verdana Corp ($142k ARR, last touch 63 days ago, health score declining).
customers.last_contact churn_signals.verdana
Feed it anything

Every source, understood.

From a single PDF to an entire Postgres database, Definable ingests, chunks, embeds, and indexes — all without you touching a line of code.

PDF
Documents
PDFs, Word, Markdown, CSVs. Drag, drop, done.
WEB
Websites & URLs
Crawl domains, single pages, or sitemaps on a schedule.
DB
Databases
Postgres, MySQL, BigQuery, Snowflake — read-only, live.
GIT
Codebases
GitHub, GitLab, Bitbucket. Branch-scoped indexing.
DRV
Cloud drives
Google Drive, Dropbox, OneDrive — folder-level sync.
WIKI
Wikis
Notion, Confluence, Coda. Workspaces or sub-trees.
TIX
Tickets
Zendesk, Intercom, Jira — resolved threads as answers.
API
Custom APIs
Bring your own source via webhook or scheduled fetch.
How a brain is built

From raw data to grounded answers.

01 INGEST
Connect your source
OAuth, API key, or drag-drop. Incremental sync from day one.
02 CHUNK & EMBED
Semantic chunking
Structure-aware splits · multi-vector embeddings · bm25 hybrid.
03 RETRIEVE
Hybrid retrieval
Vector + keyword + graph. Reranked per query.
04 ANSWER
Grounded & cited
Every claim linked to the source paragraph or row.
Grounded · always

Answers with receipts.

Every sentence is linked to the source paragraph, row, or file. No hallucinated policies, no made-up prices. When the brain doesn't know, it says so.

  • Paragraph-level citations on every answer
  • Confidence scores + "no answer" detection
  • Visual diff when source content changes
Q: What's our refund policy for enterprise?
Enterprise contracts include a prorated refund within the first 30 days [1]. After day 30, refunds require signed termination with 60 days notice [2].
[1]MSA_v3.pdf · §4.2 Refund Terms · p.14
[2]Enterprise-Policy.pdf · §9 Termination · p.22
Live sync · always fresh

A brain that keeps learning.

Connectors re-index on a schedule you set — hourly, daily, on-change. Your knowledge base mirrors reality, never the snapshot from six months ago.

  • Incremental re-embedding (only what changed)
  • Webhooks for real-time push updates
  • Version history & rollback
Sync activity · last 24h
● Notion · 12 pages updated2m ago
● Postgres · 284 new rows18m ago
● GitHub · 3 files changed1h ago
● Zendesk · 42 resolved tickets3h ago
● Drive · 7 docs synced6h ago
● Web crawl · 31 urls fetchedrunning
next scheduled run · 18:00 ISTavg re-embed · 340ms/doc

Enterprise by default.

Tenant isolation
Each brain is scoped, encrypted, and sandboxed. Zero leakage between workspaces.
No training, ever
Your content never leaves your tenant. Not used to train models.
Granular access
Per-document ACLs. Honor your existing permissions from Drive, Notion, SharePoint.
Any embedding model
Bring your own: OpenAI, Cohere, Voyage, Jina, or open-source.

Build a brain in minutes.

Connect one source. Ask a question. Get cited answers. Free on every plan.

Create your first brain Back to overview