Definite On-Prem

Stop choosing between modern and secure.

The full analytics platform and a real AI data agent, all inside your own cloud.

Single-tenant · One Helm chart · See exactly what runs where

Your Kubernetes cluster● Single-tenant
API servicesingle-tenant, isolated to you
Lakehousescalable storage, blazing-fast queries
Job runnerschedule anything, on any cadence
Fi, the AI data agentqueries, builds, and automates
§ 01 — What it is

You stop choosing between modern tooling and control of your data.

Most analytics platforms force a choice: modern tooling, or control of your data. Definite On-Prem refuses the tradeoff. You get the complete platform, ingestion, lakehouse, semantic layer, automations, data apps, and a real AI agent, running single-tenant inside infrastructure you own and your security team already trusts.

Under the hood: one Helm chart onto a Kubernetes cluster you control, a Postgres database you provision, and analytical data in your own object store. It is a purpose-built single-tenant distribution, dedicated entirely to your organization. Definite ships the software; you run it.

Regulated and security-conscious teams

You cannot send sensitive data to a third-party SaaS. With On-Prem, you do not have to.

Data-residency requirements

Deploy in the EU, a specific US region, or anywhere else. You provision the cloud resources, so the region is entirely your choice.

Teams replacing a fragmented stack

Consolidate warehouse, ingestion, BI, semantic layer, and AI tooling into one platform you run yourself.

§ 02 — Deployment

If your team runs Kubernetes, you can run Definite.

There is a standard Helm chart underneath, so nothing about this deployment is a black box: your team can audit it, script it, and fold it into existing IaC, with no lock-in. A Rust CLI (definite) drives it for convenience, but every action is also expressible as raw helm and kubectl, and definite export-helm dumps the chart for your own pipeline. Definite runs on any conformant cluster. Tell us your cloud and we will have you live on it fast.

Google Cloud
GKE · Cloud SQL · Cloud Storage
Supported
AWS
EKS · RDS · S3
Supported
Azure
AKS · Azure Database · Blob Storage
Supported
Vanilla K8s
Kubernetes 1.28+
Supported · any conformant cluster
Deploy flow · the same CLI handles status, logs, and upgrades
$ definite doctor
Preflight: probes Postgres, Kubernetes, object store, OIDC, LLM
$ definite bootstrap
Ingress, cert-manager, sandbox CRDs
$ definite init
Stand up the platform
$ definite upgrade
Pull the new chart, migrate, rolling restart, on your timeline
Runs in locked-down networks

Egress-restricted environments are a first-class deployment target, not an edge case. Mirror container images to a private registry and route Fi to an LLM endpoint inside your own VPC. If your network is locked down, Definite still runs in it.

§ 03 — Security & data control

No standing access. No copy of your data. No phone home.

Everything that touches your data runs inside your environment. Definite ships release artifacts, not a service that reaches into your cluster. Here is exactly what runs where.

01
Runs in your cloud
your Kubernetes cluster
  • API service (single-tenant FastAPI)
  • Frontend web UI
  • Lakehouse compute: DuckDB + Quack + DuckLake
  • Job runner (Postgres-backed scheduler)
  • Fi, the AI data agent, one Kubernetes sandbox per thread
02
Yours, never held by Definite
referenced by the software
  • PostgreSQL 15+ (application metadata: users, models, job state)
  • Object store: S3, GCS, or MinIO (lakehouse data, open Parquet)
  • Your identity provider (SSO via Okta or Microsoft Entra)
  • An LLM endpoint you choose at install (Anthropic API, Bedrock, Vertex, or Azure OpenAI)
03
Definite manages
release artifacts, not runtime
  • Versioned container images and the Helm chart
  • The definite CLI: preflight, deploy, upgrade, diagnostics
  • New releases on a published cadence; you apply them when you choose
  • Deployment guidance and upgrade support
  • Ed25519-signed license keys, offline-verifiable
Built so your security review is short
01

Your secrets stay yours. Database, object store, LLM keys, and license are referenced via Kubernetes Secrets, mounted files, env vars, or workload identity / IRSA, never inlined and never sent to Definite. Integration secrets are encrypted at rest with rotatable Fernet keys.

02

Off by default. Definite support access does not exist until you grant it: customer-toggled, time-bounded, and validated by a signed token.

03

Fully auditable. Every mutation is written to a filterable audit log your team owns.

04

Disable it entirely. Telemetry is a daily anonymized snapshot that fails soft when offline, and you can switch it off completely.

§ 04 — What's included

The full platform. Not a stripped-down edition.

On-Prem is the same product as Definite Cloud, not a degraded version of it. Every capability below is built, deployed, and running in the product today. None of it is roadmap.

01
Lakehouse & query
  • DuckDB + Quack + DuckLake lakehouse. Parquet data on your object store.
  • SQL query API and editor, with data-access enforced and every run logged.
  • Saved queries and recent runs, per user.
  • Data catalog: tables, schema, row preview, activity history.
  • Table freshness and SLA tracking, with per-table targets and freshness badges.
  • Lakehouse maintenance: compaction, snapshot expiry, orphaned-file cleanup.
  • Vector search and embeddings (bge-small-en-v1.5, 384-dim).
02
Data ingestion
  • Postgres-to-lakehouse sync (pg_sync), incremental and full-refresh.
  • Python ingestion steps (Polars in and out).
  • File upload: CSV, Parquet, JSON, NDJSON.
  • Integrations framework with an OAuth bring-your-own-app flow for connecting sources.
  • SSH tunnel to Postgres sources behind a bastion.
03
Semantic layer
  • DuckDB-native semantic models in Cube-format YAML, with a CRUD API.
  • Dimension and measure search.
  • Model-reference resolution: SQL can reference semantic models directly.
04
Automations
  • Linear automation pipelines with seven step types: sql, python, pg_sync, slack_webhook, agent, maintenance, llm_decision.
  • Timezone-aware cron scheduling and event triggers.
  • Conditional step guards: skip on rows_empty, succeeded, output_truthy, and more.
  • Stored, reusable Python scripts.
05
Autonomous monitoring agents
  • Probe-to-decision-to-action monitors: a SQL or Python probe, an LLM decision, an action.
  • Action modes: simple (Slack, webhook, SQL) or a full Fi agent run.
  • Cooldown suppression and cross-run memory.
06
Fi, the AI data agent
  • Conversational agent with one isolated Kubernetes sandbox per thread.
  • Around 21 built-in tools: SQL, catalog, integrations, docs search, memory, panels, web.
  • Two-layer memory: episodic run logs plus durable workspace facts.
  • Rich panels (tables, charts, images, HTML, credential forms) and live SSE streaming.
  • Reachable from the web UI, the CLI, automation agent steps, and an MCP server.
  • Optional kernel-level sandbox isolation via gVisor or Kata.
  • A concurrency guardrail caps simultaneous agent threads.
07
Data apps & embedding
  • React data apps: scaffold, build, and upload via the CLI.
  • External embedding via iframe with signed durable and short-lived tokens.
  • Per-embed row and column scoping.
08
Auth, permissions, operations
  • App roles (viewer, editor, admin) plus custom roles.
  • Per-item content ACLs across apps, threads, queries, automations, agents, integrations.
  • Table-level data-access roles with schema and table grants.
  • Long-lived, admin-issued API tokens.
  • Full audit log, inbox alerts (pipeline failure, SLA breach, schema change, Fi proposal).
  • SSO via Okta and Microsoft Entra; local Postgres auth as the fallback.
09
Web UI
  • Full web UI: Home, Query, Catalog, Fi, Automations, Agents, Integrations, Data Apps, Projects, Inbox, Docs, Settings.
  • In-app product docs and a global command palette.
§ 05 — Sources

Connect every source you already use.

The job runner connects to any source you can reach over an API, so the data you already pull into Definite comes in the same way on-prem. It runs inside your environment, with an integrations framework, Python ingestion steps, and a generic webhook to land anything else.

On-Prem ships an early integration catalog with stubs for the sources below and others. They are examples of what the job runner connects to, not a fixed list.

Salesforce
Salesforce
PostgreSQL
PostgreSQL
Stripe
Stripe
HubSpot
HubSpot
Shopify
Shopify
Snowflake
Snowflake
MySQL
MySQL
MongoDB
MongoDB
Google Analytics
Google Analytics
Segment
Segment
NetSuite
NetSuite
Intercom
Intercom
+
and any other API

If you can reach it over an API, the job runner can pull it. Postgres syncs incrementally, Python steps handle anything bespoke, and the generic webhook lands event streams, all without your data leaving your network.

§ 06 — Bring your own agent

The governed data layer for every agent in your company.

Fi is built in, but you are not locked into it. Definite On-Prem ships a read-only MCP server, so any MCP-compatible client connects and queries your governed data, all inside your environment: Claude, Cursor, an internal copilot, or a custom agent your team is building. Every one of them reads through the same data-access roles and audit log that govern your users.

01
Any MCP client

Claude, Cursor, internal and custom agents connect over the open Model Context Protocol.

02
In your environment

Agents reason over your data without it leaving your network.

03
Read-only by design

Agents query and analyze; they do not mutate your data or configuration.

04
Governed and audited

The same data-access roles and audit log that govern your users govern every agent query.

As your team adopts more agents, Definite is the one place they all reach data safely, no new copy, no new credential, no new path out of your network.

§ 07 — Architecture

Bring your own cloud.

Everything in the data path runs inside your Kubernetes cluster, in your cloud account or datacenter. The arrows from Definite into your cluster are one-time or customer-initiated: pull an image, run an upgrade, grant a time-boxed support session. There is no standing inbound connection.

Definite (vendor)
Manages release artifacts. Never accesses your data.
Container registry
versioned images
Helm chart + definite CLI
deploy / upgrade
Release cadence
published versions
Support
time-boxed · you grant access
┄┄ one-time / customer-initiated ┄┄►
Your cloud account / datacenter
your data stays in this boundary
Your Kubernetes cluster (1.28+)
API service
single-tenant FastAPI
Frontend
web UI
Lakehouse compute
DuckDB + Quack + DuckLake
Job runner
Postgres-backed scheduler
Fi
AI data agent
PostgreSQL 15+
app metadata
Object store
S3 / GCS / Azure / MinIO
OIDC provider
your IdP · optional
Your data sources
ingested or pushed
LLM endpoint
Fi connects to a managed LLM you choose: Anthropic, Bedrock, Vertex, or Azure OpenAI. It is the only component that can sit outside your network, and even that is optional. Point Fi at a Bedrock or Vertex endpoint inside your VPC and your data stays in your environment. Fi sends prompt context, never bulk data exports.
Solid borderin-environment data path, always inside your cloud
Dashed borderoptional or customer-initiated; no standing inbound connection
§ 08 — Pricing

One number a year. No metering, no surprises.

On-Prem Standard
From $48,000/ year
  • Single-tenant deployment in your cloud
  • One Helm chart, the definite CLI, versioned releases
  • The complete Definite analytics platform
  • Hands-on deployment support and an operations runbook
Schedule a deployment call
Enterprise
Contact sales
  • Higher support tier and SLAs
  • We complete your security review and vendor questionnaires
  • Scoping tailored to your environment and data volume
Schedule a deployment call
Design Partner program
First wave· 5 slots
  • Discounted first-wave pricing
  • Deploy alongside Definite's team, with direct founder access
  • Direct input into what we build next
Schedule a deployment call

On-Prem buyers typically prefer a flat annual contract over metered usage. We'll scope the right plan with you on the deployment call.

§ 09 — FAQ

Questions from security teams, answered.

Definite (the company) is SOC 2 Type II certified, with passed audits and penetration tests. Review the full posture at trust.definite.app.

No. All ingestion, storage, query execution, and dashboard rendering happen inside your environment. Analytical data lives in your object store in open Parquet format; application metadata lives in your Postgres. Definite has no copy and no access path.
Yes. One deployment equals one organization. No shared infrastructure, no shared database, no multi-tenant partitioning. You are the only tenant.
Nothing in your environment by default. No standing access, no hidden network path out, no phone-home that transmits your data. For support, you can optionally grant time-boxed, audited access. The deployment requires no inbound connection from Definite.
Yes. On-Prem supports SSO via Okta and Microsoft Entra (OIDC). Local Postgres-backed auth is available for initial setup and as a fallback. We will scope coverage to your provider during onboarding.
Secrets (database passwords, object store credentials, LLM API keys, license key) are referenced, never inlined. The config supports Kubernetes Secrets, mounted files, environment variables, or workload identity / IRSA so you can avoid static long-lived keys entirely. Definite never receives any of these secrets.
Fi reasons over your data using an LLM you provide credentials for. The data Fi sees stays in your environment; only the prompt context Fi constructs reaches the LLM endpoint you chose. Point Fi at a Bedrock or Vertex endpoint inside your VPC and nothing leaves your network.
Customer-initiated. definite upgrade pulls the new chart version, runs database migrations, and performs a rolling restart. You choose when; upgrades are never forced. The deployment is a standard Helm chart, so SREs can manage it with their own tooling.
Yes. We complete standard security questionnaires and join calls with your InfoSec or vendor-risk team. The On-Prem model usually shortens these reviews: the highest-risk questions (where does our data go, who can access it) have a simple answer: it stays with you. Definite (the company) is SOC 2 Type II certified; our Trust Center is at trust.definite.app.
§ 10 — Early access

Five deployment slots in the first wave.

If your team has been blocked by security, compliance, or data-residency requirements, this is the version built for you.

We are taking five deployments in this first wave so we can work hands-on with each platform team through install, security review, and first data load. You get direct founder access and a fast deployment.

Schedule a deployment call with Mike
First-wave deployment slots● Open
01On-prem deployment
02On-prem deployment
03On-prem deployment
04On-prem deployment
05On-prem deployment

Keep your data in your cloud.

In 30 minutes with Mike Ritchie, founder of Definite, you will have answers for your security team, a deployment scoped to your environment, and a clear path to live. He will walk your team through the architecture and answer your security questionnaire on the call.