Enterprise Decision Intelligence

Decisions you can prove, not dashboards you have to trust.

Atlas Dig reads your AWS, Azure, and GCP accounts — read-only — and turns them into a knowledge graph with evidence behind every fact. Ask what depends on something. Decide if it's safe to remove. Get a deterministic answer, with the proof attached.

Self-hosted, single-tenant Read-only by contract No black-box LLM verdicts
atlas — decide remove
$ atlas discover aws-organization --mode auto --database-url $DATABASE_URL
→ 3 accounts, 214 assets synced, 189 relationships built
 
$ atlas decide remove --entity "payments-api-ec2"
→ verdict: CONDITIONAL
  2 active dependents · 1 runbook reference · last traffic 4h ago
  evidence: sg-0a2f... → vpc-flow-logs (fresh, 11m) · route53 → payments-api.internal
 
$
The problem

Your infrastructure record and your infrastructure disagree.

Three tools claim to know what's running. None of them can tell you what happens if you touch it.

01 — CMDB

A snapshot, not a system of record

It's accurate the day someone updates it, and wrong every day after. Nobody re-syncs a spreadsheet under deadline pressure.

02 — DASHBOARDS

State without consequence

A graph of resources tells you what exists. It doesn't tell you what breaks if you remove one, or who's still relying on it.

03 — LLM COPILOTS

Confident, fluent, and sometimes wrong

A plausible-sounding answer with no evidence chain is worse than no answer — you find out which, in the incident review.

How it works

From raw cloud APIs to a decision with a paper trail.

1

Discover

Read-only connectors pull real state from AWS, Azure, and GCP. Every connector declares ReadOnly: true — Atlas Dig refuses to run one that doesn't.

2

Build the graph

Assets and relationships land in a Universal Knowledge Model. Every fact carries provenance, a freshness window, and a trust score.

3

Ask, or decide

Query in plain language, or run a real decision — remove, migrate — and get YES, NO, CONDITIONAL, or UNKNOWN. Never a guess dressed up as one.

4

Prove it

Every answer carries its evidence chain and lands in an append-only audit trail — the record you hand to the reviewer, not the one you reconstruct afterward.

What you get

Built for the team that has to answer for the decision.

Multi-cloud, read-only

AWS Organizations, Azure subscriptions, GCP projects — one command scans a whole org via cross-account AssumeRole, never standing credentials.

Provenance on every fact

No node in the graph exists without a source and a timestamp. Stale evidence is labeled stale, not silently trusted.

Deterministic decisions

remove returns YES, NO, CONDITIONAL, or UNKNOWN — the same input always produces the same verdict, with its reasoning attached.

Self-hosted, single-tenant

Runs inside your own infrastructure. Your inventory, evidence, and decisions never leave it — nothing phones home unless you opt in.

RBAC, SSO, and an audit trail

Role-based access, OIDC/SSO for enterprise identity, and an append-only log of every decision — built for the security review, not just the pilot.

Licensing that never breaks a build

A signed, offline license key — no license server. If a renewal lapses, only new discovery pauses; everything you've already found keeps working.

Security & trust

Built to pass the security review, not just the pilot.

Most tools ask for standing access and hope you don't read the IAM policy too closely. Atlas Dig is built the other way around: least privilege first, evidence for every claim, and nothing that leaves your environment unless you explicitly turn it on.

  • Read-only by contractEvery connector declares ReadOnly: true at compile time; Atlas Dig refuses to run one that doesn't.
  • Cross-account AssumeRole, least privilegeEvery account is discovered through a role you grant, scoped to read-only. No long-lived credentials, ever.
  • Self-hosted, in your VPCYou run the containers. Your data stays on your infrastructure — there's no multi-tenant database to worry about.
  • Signed, offline license verificationNo license server, no outbound call to prove you're entitled to run it. Usage reporting is opt-in and previewable before you turn it on.
Pricing

Priced by what you actually run, not by seats.

Annual, billed in advance, on the daily average of discovered assets — not the peak. Business-context entities you add yourself are never billed.

Starter
$4,800/year
Up to 2,000 assets · $2.40/asset/year
  • AWS, Azure & GCP discovery
  • Knowledge graph & evidence trail
  • Deterministic remove decisions
  • Self-serve checkout
  • Email support
Start free trial
Enterprise
Quoted
10,000+ assets
  • Everything in Growth
  • Volume-based pricing
  • Deployment & onboarding support
  • Direct line to the founder
Talk to us

"Discovered asset" means an entity that came from a real discovery run. Teams, applications, capabilities, and runbooks you add to describe your own organization are never counted or billed.

Try it on your own infrastructure.

30 days, no credit card. The trial clock starts on your first successful discovery run — not the day you sign up.