Terraform names your pets. Terrantula herds your cattle.
Visibility, cascade ordering, and tenant-fleet orchestration — on top of the Terraform runner you already run.
Sits on top of Terraform
The cattle wedge
Terraform names your pets. Terrantula herds your cattle.
Terraform is excellent for pets — long-lived infrastructure you curate by hand. It's awkward for cattle: fleets of templated, often-ephemeral resources where the population matters more than the individual. Terrantula adds the missing layer above your runner.
Visibility graph
A single fleet view across every tenant, cluster, and stack — derived from the Terraform state you already produce. No more state files scattered across N workspaces with no system-level picture.
Cascade ordering
Typed relationships with cardinality and lifecycle. Terrantula computes the order things must change in — placement, dependency, deprovisioning — so the fleet stays consistent as it grows.
Tenant-fleet orchestration
Define a tenant blueprint once. Terrantula handles placement onto the least-loaded matching cell, capacity enforcement, and lifecycle across hundreds of tenants — using your existing runner.
See it
The fleet, as a graph
The Web (Strand) graph view renders your fleet as a living graph — entities, cells, and relationships. It is a read-only projection of your Terraform-derived state: the view never writes infrastructure. All changes flow through Actions → pull requests → your apply.
What's in the box
Four primitives, one orchestration model
The model SaaS operators already build by hand, every time — made first-class.
Entities & Cells
Typed, stateful records of what exists — tenants, clusters, databases — placed into cells with capacity constraints and a placement policy.
Relationships & Cascade
Model dependencies with cardinality enforcement. Terrantula derives the cascade order for changes and deprovisioning from the graph.
Actions → PRs
Actions modify your IaC by opening a pull request against your repository. Terrantula never runs terraform apply — your CI applies the change, exactly as it does today.
Self-hostable
Apache-2.0 backend. The cattle wedge runs end-to-end on free, self-hostable tooling — Terrantula + your CI + your state, with zero required SaaS dependencies.
How it works
From state file to fleet, in five steps
The first five minutes get you a working graph view against your existing state. Everything after that flows through pull requests into the runner you already run.
Import your Terraform state
Point Terrantula at a .tfstate file (or an Atmos stacks directory) and a starter config. It maps your resources to typed entities — no blueprint required.
Learn more: Import your TF state →terrantula import-terraform \ --state ./terraform.tfstate \ --config aws-multi-accountOpen the dashboard
Launch the local fleet view. It runs entirely on your machine — no signup, no server, no Docker. The graph is a read-only projection of your Terraform-derived state.
Learn more: Quick Start (5 minutes) →terrantula dashboard # Dashboard ready at http://localhost:54321Model cells & relationships
Describe the fleet in YAML: entity types, cells with capacity constraints, and typed relationships with cardinality. This is the structure the cascade reasons over.
Learn more: Core Concepts →kind: Cell metadata: name: prod-clusters spec: entityType: TenantCluster placementPolicy: least-loaded constraints: - { metric: tenant-count, aggregate: sum, max: 500 }Run an Action — it opens a PR
An Action like OnboardTenant places the tenant onto the least-loaded matching cell and opens a pull request against your IaC repository. Terrantula never applies it for you.
Learn more: Wire an Action →kind: Action metadata: name: OnboardTenant spec: associatedWith: { entityType: Tenant, scope: collection } recommendations: - { name: target-cluster, pool: prod-clusters, sortBy: tenant-count }Your CI applies the change
The PR lands in your existing workflow. Whatever runs Terraform for you today — TFC, Atmos, Atlantis, or a GitHub Actions workflow with state in S3 — keeps doing it, exactly as before.
Learn more: Advanced Quick Start →# Reviewer merges the PR → # your existing pipeline runs the apply. # Terrantula syncs the result back into the graph.
Self-hostable, end to end.
The backend is Apache-2.0. The cattle wedge runs on free, self-hostable tooling — Terrantula + your CI + your state — with zero required SaaS dependencies. Start local in under five minutes.