Governed analytics platform
Build once.
Deliver analytics everywhere.
Datafor unifies self-service BI, embedded analytics, and AI-powered analysis on one governed semantic model. Every dashboard, customer portal, and AI answer uses the same numbers and the same permissions.
NewDatafor 10.00: MCP access for Claude and ChatGPT, metrics from models, rebuilt filters →Self-service BI
- Dashboards
- Reports
- Drill-down
Embedded
- SDK & URL embed
- White-label
- Multi-tenant
AI
- AI Agent
- MCP clients
- Insights
Governed semantic model
Trusted by teams that ship analytics to their users
- Hitachi
- GEP
- DerbySoft
- Aspire
- OptimusFleet
Why analytics stops scaling
Three problems appear when analytics moves beyond dashboards
Inconsistent metric definitions
Different versions of the same metric create conflicting reports, rework, and lower trust in decisions.
Embedded analytics rebuilt per product
Analytics gets rebuilt for every product or customer, making reuse, OEM delivery, and multi-tenant scale expensive.
Ungoverned AI answers
Natural-language analytics is going mainstream, but reliability, traceability, and permission guardrails are often missing.
One foundation
Three delivery experiences, one set of definitions
Model metrics, relationships, and access rules once. Every surface reuses them.
For internal teams
Dashboards, reports, and exploration
Build dashboards on the shared model, drill from summary to detail, and export reports that keep their formatting.
Learn more →
For product teams & ISVs
Embedded analytics inside your product
Embed reports with the SDK or URLs, apply your brand, and isolate every tenant without rebuilding analytics per customer.
Learn more →
For AI-enabled workflows
Natural-language analysis with guardrails
Ask questions in Datafor, Claude, ChatGPT, or Cursor. Answers come from the governed model, under the asker's own permissions.
Learn more →
AI you can verify
The AI never writes SQL. Datafor does, under your rules.
The language model only proposes a structured query from your model's metadata. Datafor validates it and runs it with the asker's permissions, so an answer can never see more than the person asking.
- 1
Ask in plain language
In the Datafor AI Assistant or any MCP client.
- 2
The AI proposes a structured query
Built from model metadata: metrics, dimensions, time fields. The AI never writes SQL or MDX.
- 3
Datafor validates and runs it
Checked against the model and executed under the user's row-level security.
- 4
You see the rows and the query
Every answer carries its result, the query behind it, and any disclosures.
Bring Datafor into Claude, ChatGPT, or Cursor
Since 10.00, any MCP client can query Datafor with a personal token. Read-only, permission-aware, revocable.
claude mcp add --transport http datafor "<MCP address>" \
--header "Authorization: Bearer <token>"Governance across every surface
Security that travels with the data
The same policies apply in dashboards, embedded reports, APIs, and AI answers.
Row and column security
Condition-based row policies plus table and column hiding, enforced on every surface.
Test access
Simulate what any user or role can see before you roll a policy out.
Certified metrics
A Metrics Library with Draft and Certified status, stable IDs, and definition drift checks.
Personal tokens for AI clients
MCP access uses per-user, expiring, revocable tokens that admins control.
Tenant-wide settings
Defaults per tenant with conflict checks and an audit trail.
Bring your own LLM
Provider presets, per-stage model assignment, usage quotas, and keys that are never exported.
See Datafor on your own data
Book a guided session for your scenario, or start a 30-day trial on your own infrastructure.