Terraform Cloud alternatives and Terraform automation tools, compared

Terraform automation and collaboration software, often called TACOS, runs plan and apply for a team: on pull requests, with locks, approvals and a record of what ran. These tools differ most on three things: where Terraform runs, who holds your cloud credentials, and how they order dependent stacks. The matrix below compares Stackorder with 10 other tools on those and 14 other features, with a source for every cell.

Stackorder is an open-source orchestrator for Terraform and OpenTofu on GitHub Actions. Its server never runs Terraform and needs no cloud credentials; your GitHub Actions runners do the work.

Updated . Last reviewed ; Terramate and Terragrunt against each tool's own repository, documentation or pricing page.

Feature matrix

A dash means we haven't verified it, not that it's missing. Scroll sideways to see every tool; each tool's page cites every cell.

FeatureStackorderAtlantisHCP Terraform (formerly Terraform Cloud)Stategraph (formerly Terrateam)Spaceliftenv zero (formerly env0)ScalrTerrakubeOpenTaco (formerly Digger)TerramateTerragrunt
LicenseApache-2.0, open sourceApache-2.0, open source; a CNCF Sandbox projectProprietary, commercial; the Terraform CLI itself is under BSL 1.1 since version 1.6.0MPL-2.0; Enterprise features need a commercial licenseProprietary platform; open-source tooling such as spacectl and the Terraform provider under MITProprietary SaaS; its Terraform provider, Terratag and MCP server are open sourceProprietary SaaS; its Terraform provider is MPL-2.0 and its GitHub Action MITApache-2.0, open sourceOpen core: MIT at the root, and a proprietary Digger Enterprise license for the ee/ directoryMPL-2.0 for the CLI; Terramate Cloud is a commercial SaaSMIT
DeploymentSelf-hosted; setup mode creates the GitHub App from a manifestSelf-hosted only: Helm chart, Kubernetes manifests or Kustomize, OpenShift, an AWS Fargate module, GKE or GCE, or DockerSaaS, with a separate HCP Europe region; Terraform Enterprise is the self-hosted distributionStategraph Cloud (SaaS), self-hosted (open source or Enterprise), or in your own AWS, GCP or Azure accountSaaS, now called Spacelift Deploy; self-hosted on AWS, Azure, GCP or on-premises Kubernetes only on the Enterprise+ tierSaaS only, optionally with self-hosted agents; the control plane cannot be self-hostedSaaS only, at scalr.io on Google Cloud in the United States; agents can run jobs in your networkSelf-hosted only: a Helm chart on Kubernetes, or Docker ComposeHosted app at otaco.app, or self-hosted with Docker Compose, Railway or Helm chartsThe CLI runs in your CI or shell; Terramate Cloud is SaaS, or a Private Cloud instance on EnterpriseA CLI you run yourself; Terragrunt Scale runs in your GitHub Actions or GitLab CI
PricingFree and open source; you run the serverFree; no paid tier or hosted offeringPer managed resource, billed hourly: Free up to 500 resources; Essentials from $0.10, Standard from $0.47 and Premium from $0.99 per resource a month; Terraform Enterprise customCloud Free: 3 users, 50 runs a month, 1 worker. Pro: $12,000 a year with unlimited users, runs and workers. Self-hosted open source: $0, up to 3 active users a month per installationFree: 2 users, 1 public worker. Starter+: $20,000 a year, unlimited users. Business, Enterprise and Enterprise+ by quoteFree: 250 runs a month, 30 active environments, 1 concurrent run, 1 self-hosted agent. Cloud Navigator and Cloud Pilot: custom, priced per successful apply or environmentPer run. Free: 50 runs a month with unlimited users, SAML SSO and agents. Business: $0.99 a run. Enterprise: custom, from 20,000 runs a yearFree; no paid tier, hosted service or commercial edition advertisedNo public pricing page; Enterprise features by per-seat subscription, through contact or a demoCLI free. Cloud Community: free, 2 users, 1,000 resources. Teams: $449 a month, unlimited users, 5,000 resources. Enterprise: customCLI free. Terragrunt Scale: free up to 25 units; Team from $500 a month for 200 units, on an annual contract; Enterprise custom
Maturityv0.6.0, released 2026-11-15; first released as v0.1.0 on 2026-09-30A CNCF Sandbox project, actively released; v0.48.0 added slim images—The vendor reports daily releases and more than 20,000 automated tests per change———Actively maintained: 2.33.2 released on 2026-09-14, with 2.34.0 in betaActively released: v0.6.151 on 2026-09-13, after roughly monthly releases through 2026Actively released: v0.17.3 on 2026-09-08, four releases since May 2026v1.2.1 on 2026-10-23; 1.0 on 2026-03-30, with no breaking changes to the CLI or HCL in 1.x minor releases, experimental features excepted
Where Terraform runsYour GitHub Actions runners, GitHub-hosted or self-hosted; it manages no runners or agentsOn the Atlantis server itself, not on CI runnersDisposable VMs in HashiCorp's cloud by default, or self-hosted agents that need only outbound access; a workspace can also run locallyYour GitHub Actions or GitLab CI runners; the server dispatches workflows and does not run TerraformA fresh Docker container per job, on Spacelift's public workers or on private workers you runenv zero-hosted agents by default, or self-hosted Kubernetes or Docker agents in your infrastructureScalr's shared runners by default, or any number of self-hosted agents, each adding 5 concurrent runsIts own executors: a persistent pod pool, ephemeral Kubernetes Jobs, or self-hosted agentsYour CI, GitHub Actions by default, for pull request automation and drift; Remote Runs (beta) in E2B sandbox VMsWherever the CLI runs: it runs any command in each selected stack, in your CI or locally; Terramate Cloud runs nothingWherever you run the CLI; Terragrunt Scale runs it on your GitHub Actions or GitLab CI runners
State backendBring your own S3; never takes or releases the state lockBring your own; any backend except local stateBuilt in: HCP Terraform is the state backendYour existing backend; Stategraph's database-backed state is a separate commercial product. Plan files stay in the server's database by defaultOptional Spacelift-managed state on S3, with history and rollback, chosen when a stack is created; or your own backendOptional env zero remote backend with locking and versioning, which can store state in your own S3 bucket; or a standard backend such as S3Built in, in a Scalr-owned encrypted bucket; your own bucket through storage profiles on Enterprise; other backends with limitsBuilt in, per workspace, in Azure Storage, S3, Google Cloud Storage or MinIO; implements the TFE remote backend and cloud blockYour existing backend; or OpenTaco's HCP Terraform-compatible state service on S3-compatible storageYour existing backend; Terramate Cloud needs no access to itAny backend OpenTofu or Terraform supports; remote_state generates the backend block and, with --backend-bootstrap, can create S3, GCS or Azure storage
ModulesNo registry; lists each module's consumers, and for git modules the version each pins and how far behind it isNo registry; the opt-in --autoplan-modules plans the projects that use a changed local moduleBuilt-in private registry versioned by Git tags; the Explorer shows module usage across workspacesNo registry; a module-aware indexer, off by default, plans the directories that use a changed local modulePrivate module registry on paid accounts, with tests per version; module trigger policies see each module's consumersPrivate module and provider registry, with download counts and the environments using each moduleBuilt-in private registry, published from VCS or OCI; a Modules report shows versions across workspaces and flags outdated onesBuilt-in private registry for modules and providers; a new SemVer tag imports a module versionNo registry; include_patterns globs trigger projects when a module path changesNo registry; change detection follows local modules referenced by relative path, through any depth of nestingNo registry; each unit's terraform.source points at a module, with module source syntax
Self-hosted footprintOne container of about 30 MB and Postgres; actions that use no DockerOne Go binary or container with no external database; a persistent disk for plans and BoltDB locks, or Redis for locksSaaS. Terraform Enterprise runs as containers with PostgreSQL, S3-compatible storage and Vault, plus Redis for active-activeServer and PostgreSQL behind a public HTTPS URL; a GitHub Action that runs as a Docker containerSaaS. Self-hosted: server and drain services on Kubernetes or ECS, PostgreSQL, object storage buckets, a queue and an optional MQTT brokerSaaS control plane. Kubernetes agent: Kubernetes 1.24 or later, one pod per deployment requesting at least 460m CPU and 1500Mi memorySaaS control plane. Agent: at least 2 GB of memory and 1 CPU per run, 20 GB of disk, and outbound HTTPS to scalr.ioAPI, executor, registry, UI, Dex and OpenLDAP (on by default), with MinIO, PostgreSQL and Valkey subcharts that can be swapped for external servicesUI, orchestrator, drift, drift trigger, state, token and sidecar services; three Postgres databases, S3-compatible storage and WorkOSNo server to self-host: the CLI in your CI, with Terramate Cloud as SaaSNo server to run: the CLI, and Pipelines as a binary in your CI; Gruntwork hosts the Gruntwork.io GitHub App and the developer portal that manages it
Cross-stack dependenciesA graph of stacks and modules from depends_on, module sources, and terraform_remote_state and stackorder_outputs reads, including cross-repository edges; applies in wavesexecution_order_group for a global plan or apply, and depends_on between projects in one atlantis.yamlRun triggers between workspaces, up to 20 sources each; Stacks, with linked Stacks, on RUM plansLayered runs: depends_on between directories in one repository, optionally on specific outputsStack dependencies form a DAG, across repositories, and pass outputs downstream as inputsWorkflows, in which each environment lists what must deploy first under needs; Workflow Triggers chain deploysExplicit run triggers between workspaces, up to 50 upstreams, also across federated environments; not inferredStable 2.33: shared state through terraform_remote_state. Run triggers with a dependency graph only in 2.34.0 pre-releasesdepends_on between projects and numeric layers in digger.yml; Terragrunt dependency parsingParents run before children, and after and before order stacks; these order a run but add no stacks to it. --include-all-dependents adds stacks that read outputs or Terragrunt dependency blocksdependency blocks order units and expose their outputs, dependencies blocks only order them; run --all runs dependencies first, and destroys in reverse
Sharing outputs between stacksShared outputs, off by default: the producer lists its readers in outputs.readers, and each reader uses the stackorder_outputs data source or ephemeral resource. The server returns only the outputs, sensitive ones only if the producer shares them, so the reader's role needs no access to the producer's state; a reader's plan fails while the server is unreachableNone built in: a project reads another with terraform_remote_state, so the server's credentials must read the whole state; ordering passes no valuesThe tfe_outputs data source, with a token that may read a workspace's outputs only, not its whole state; or terraform_remote_state with Remote State Sharing, which reads the whole state—Stack dependencies pass the upstream outputs you pick to the downstream stack as environment variables; sensitive ones only when enabled on the stack and worker pool. External state access, off by default, lets terraform_remote_state read a stack's stateEnvironment Outputs, a project opt-in: a variable names another environment's output, with no access to its state; string outputs only, with sensitive ones stored encrypted by env zero. Or terraform_remote_state against its remote backendterraform_remote_state against the Scalr backend, with a per-workspace choice of which workspaces may read; Scalr says the data source reads only outputs, not the whole stateterraform_remote_state against the Terrakube host, since 2.15.0; no sharing setting or outputs-only access documentedNone built in: a project reads another with terraform_remote_state and your CI role, which reads the whole state; layers order runs only, and OpenTaco's state RBAC documents no outputs-only grantOutputs sharing, experimental: an input block reads another stack's output at run time through a command, usually terraform output -json, which reads that stack's state; mock values cover plans before it appliesdependency.<name>.outputs, read from the dependency's remote state by default since v1.2.0, so the job needs access to that state; mock_outputs stand in before it applies
Cloud credentialsNot held by the server: each job assumes your IAM role with its own GitHub OIDC token. Shared outputs, off by default, give the server a read-only role on the state objects you listHeld by the server, which runs Terraform: instance or workload roles, environment variables, credential files or VaultHeld by the service: static credentials as variables, or short-lived OIDC credentials it receives for each runNot held by the server; the runner uses OIDC or secrets stored in GitHub or GitLabSpacelift assumes your AWS role, or a private worker does and the credentials never reach Spacelift; Spacelift-signed OIDC as an alternativeOn hosted agents, env zero authenticates to your cloud: assume-role, stored keys, or OIDC tokens it issues; self-hosted agents read your own secret storeHeld by default in encrypted provider configurations: keys, an IAM role, or per-run OIDC with Scalr as the identity provider; agents can keep them in your networkHeld by the platform: static credentials as sensitive variables, or short-lived tokens from Terrakube's own OIDC issuerNot held for pull request automation: credentials stay in your CI, with OIDC recommended; not documented for Remote RunsNot held: the CI job assumes your role, with OIDC in the blueprints; Terramate Cloud needs no access to cloud accountsTerragrunt Scale: the Pipelines binary in your CI assumes your roles through OIDC, with temporary credentials per environment
Human sign-inGitHub OAuth through the App, read:org scope only; the AWS Terraform module can also require OIDC or Cognito sign-in at the load balancer, in front of the UI———SAML 2.0 single sign-on from the Enterprise tier—SAML single sign-on, included on the free planDex connectors with groups, such as Azure AD, Google, Cognito, Keycloak, GitHub, GitLab or LDAPWorkOS, required when self-hostingTerramate Cloud: SSO on every pricing tier; SAML 2.0, RBAC and audit logs on Enterprise—
Git hostsGitHub only, by designGitHub, GitLab, Gitea and Forgejo, Bitbucket Cloud and Server, Azure DevOpsGitHub, GitLab, Bitbucket and Azure DevOps, hosted and self-managed; an API-driven workflow for othersGitHub, including GitHub Enterprise Server, and GitLabGitHub (including GitHub Enterprise), GitLab, Azure DevOps, Bitbucket Cloud and Data Center, and raw GitGitHub, GitLab, Bitbucket and Azure DevOps; self-hosted GitHub Enterprise Server, Bitbucket Data Center and GitLab through an agentGitHub, GitHub Enterprise, GitLab, Azure DevOps, Bitbucket Cloud and Data Center; VCS agents for private networksGitHub (including Enterprise), GitLab, Bitbucket, Azure DevOps, and SSH repositories without webhooksGitHub is the documented path; GitLab pipelines need an Enterprise license keyCI blueprints for GitHub Actions, GitLab CI and Bitbucket Pipelines; pull request previews sync from GitHub Actions onlyTerragrunt Scale: GitHub and GitLab; self-hosted GitHub Enterprise and GitLab on the Enterprise tier; the free tier, GitHub and AWS only
OpenTofuYes, with tool: tofu; tested end to end with OpenTofu 1.12Yes, with --default-tf-distribution=opentofu or terraform_distribution per projectNot documented; it runs the Terraform CLIYes, with engine: tofu; also Terragrunt, CDKTF and PulumiYes, first-class; all OpenTofu versionsYes; OpenTofu is the default binaryYes, first-class, billed the same as TerraformYes, chosen per workspaceYes, per project with opentofu: true; Terragrunt and Pulumi tooYes; it runs tofu like any other command, and Terragrunt tooYes; it runs tofu by default and falls back to terraform
Drift detectionScheduled per stack with drift.schedule; with open_issue, one GitHub issue per drifted stack, closed when the drift is gone; never applies to fix driftAlpha API endpoints since v0.45.0, off by default; no built-in scheduler, so an external job has to call themHealth assessments on Standard and Premium, for workspaces in remote or agent executionScheduled hourly to monthly, off by default; can open an issue or apply fixes with reconcile; one schedule in the open-source editionScheduled proposed runs with optional reconciliation; private workers only; Starter+ and aboveScheduled per environment, with alerts; optional automatic remediation by redeploying or opening a pull requestScheduled daily or weekly per environment, with ignore, sync-state or revert actions; drift runs are not billedA documented pattern: a plan template with an OPA check and a Slack notice, run on a workspace scheduleA drift service on a cron schedule, or a backendless Actions workflow, reporting to Slack or GitHub IssuesDrift workflows you schedule in your CI, synced to Terramate Cloud, with Slack notificationsTerragrunt Scale, Team and up: terragrunt plan on a schedule, opening a pull request for the drifted units
Policy checksNot a policy engine; run OPA, conftest, Checkov or Infracost in hooks, and stackorder check records a named check the apply gate honorsBuilt-in Conftest (OPA) checks with policy-owner approval; custom_policy_check for other toolsBuilt in: Sentinel, OPA, and HCL-based Terraform policy in betaOPA/Rego, Conftest and Checkov on every plan in every edition; Gatekeeper approval gates in Stategraph Cloud and self-hosted EnterpriseBuilt-in OPA/Rego policy engine with several policy typesOPA approval policies after plan and cost estimation, not enforced on pull request plans; cost estimates through InfracostBuilt-in OPA before and after plan, with three enforcement levels; Checkov before planOPA through extensions and templates in stable releases; native policy governance in 2.34.0 pre-releasesOPA through Conftest, run as a workflow step against the planTerramate Cloud: more than 500 built-in CIS Benchmark policies, run in drift checks, and policy as code with OPA—
Pull request workflowA check per stack, one sticky comment, and stackorder plan, apply and unlock comments; applies before merge by default, or on mergeatlantis plan and atlantis apply comments and autoplan on each commit; applies before merge by default; a lock per directory and workspace until the pull request merges or closesSpeculative plans posted as pull request checks; merges trigger plans; manual or automatic apply per workspaceAutoplan on each push, against the merge with the destination branch; a stategraph apply comment applies the reviewed plan; applies before merge by default, or on mergeProposed runs on pushes, reported as commit statuses; opt-in plan comments and /spacelift commands; deploying before merge is supportedPlans on pull requests as a comment and status checks; plan and apply from comments, open to any commenter unless RBAC enforcement is turned onA dry run on every pull request; /scalr plan and /scalr apply comments allow apply before merge; GitHub checksWebhooks on push, pull request or release; an optional plan comment and terrakube plan and terrakube apply commands, except on Azure DevOpsdigger plan, apply, lock and unlock comments; applies before merge by default, on merge if configured; pull request locksThe blueprints plan changed stacks on each pull request and apply them on push to main; Terramate Cloud posts a preview summary in the pull request, with the full plans in Terramate CloudTerragrunt Scale's Pipelines posts plans as pull request comments and applies in dependency order on merge

What to look for in a Terraform Cloud alternative

Where Terraform runs
On the vendor's workers, on the tool's own server or executors, or on your own CI runners.
Who holds cloud credentials
The platform, a server you run, or your CI runner, which can assume a role with its own OIDC token.
Sharing outputs between stacks
Whether a stack can read another stack's outputs without access to its whole state, which terraform_remote_state usually needs.
State and modules
Whether the tool hosts state and a private module registry, or works with the backend and Git repositories you already have.
Git hosts
GitHub only, or GitLab, Bitbucket and Azure DevOps as well.
OpenTofu
Supported as a first-class tool, chosen per project or workspace, or not documented.
Cross-stack dependencies
Run triggers between workspaces, ordering within one repository, or a graph that also follows modules and state reads.
Pricing model
Free and open source, per managed resource, per run, per seat, or by quote.

Which tool fits which team

Your code is not on GitHub, or is on several Git hosts
Atlantis, Stategraph, HCP Terraform, Spacelift, env zero, Scalr and Terrakube all work beyond GitHub; OpenTaco's GitLab pipelines need an Enterprise license key. Terramate's CLI runs in any CI, while its pull request previews sync from GitHub Actions only, and Terragrunt Scale supports GitHub and GitLab, and its free tier GitHub only. Stackorder is GitHub only, by design.
You want Terraform to run on your own CI runners, with cloud credentials in your CI
Stackorder runs on your GitHub Actions runners, Stategraph on your GitHub Actions or GitLab CI runners, OpenTaco in your CI, GitHub Actions by default, Terramate in whichever CI runs its CLI, and Terragrunt Scale on your GitHub Actions or GitLab CI runners.
You want plan and apply from pull request comments on one server, with no CI runners involved
Atlantis runs Terraform on its own server, which holds your cloud credentials.
You want to self-host a platform that stores state and hosts a private registry
Terrakube is Apache-2.0 and does both. Terraform Enterprise, the self-hosted edition of HCP Terraform, is the commercial option.
You want a policy engine built in
Spacelift, Scalr, HCP Terraform, Stategraph and env zero evaluate OPA or Sentinel policies themselves, and Atlantis runs Conftest. Stackorder records the verdicts of the tools you run in hooks.
You want a managed service with no control plane to run
HCP Terraform, Spacelift, env zero and Scalr are SaaS, and Stategraph and OpenTaco also offer hosted editions.
You want stacks run in dependency order with no server to host
Terramate's CLI orders stacks with after and before on every run, and Terramate Cloud, a SaaS, is optional. Terragrunt's run --all applies units in the order their dependency blocks set, from a laptop or any CI.
You have many stacks on GitHub that depend on each other and on shared modules, with state in S3
Stackorder plans every stack a change reaches, through module sources and dependencies, and applies them in dependency waves ordered by depends_on and terraform_remote_state and stackorder_outputs reads.

One page per tool

What Stackorder is, and what it is not

Stackorder is a GitHub App plus a small control-plane server that decides which stacks to run and in what order, then lets GitHub Actions do all of the running. By default its server does two jobs: dependency resolution, and a record of every plan, apply and drift check. It is deliberately not a state backend, module registry, secrets store, policy engine or runner.

Read the comparison in the documentation for the reasoning row by row, or the overview of how Stackorder works.

When something else fits better

  • You are not on GitHub. Stackorder is GitHub only, by design.
  • You want the tool to host state or modules. Stackorder stores neither. Stack discovery looks for backend "s3" blocks, and state links assume S3.
  • You want a policy engine built in. Stackorder records the verdicts of the tools you run; it does not evaluate policy itself.
  • You want managed runners or agents. Use GitHub's own runner mechanisms; Stackorder does not manage them.

Frequently asked questions

What is the best open-source alternative to Terraform Cloud?

For teams on GitHub with state in S3, Stackorder (Apache-2.0) is an open-source Terraform Cloud alternative that runs plans and applies on your own GitHub Actions runners: see Stackorder as a Terraform Cloud alternative. Terrakube (Apache-2.0) is the closest self-hosted match if you need the tool to host state and a private registry, and Atlantis (Apache-2.0) runs plan and apply on its own server from pull request comments.

Which of these tools run Terraform on GitHub Actions?

Stackorder runs every plan and apply on your GitHub Actions runners. Stategraph Orchestration, formerly Terrateam, runs them on your GitHub Actions or GitLab CI runners, and OpenTaco, formerly Digger, runs its pull request automation in your CI, GitHub Actions by default. Terramate and Terragrunt run Terraform wherever their CLIs run, GitHub Actions included, and Terragrunt Scale runs it on your GitHub Actions or GitLab CI runners. Atlantis runs Terraform on its own server, HCP Terraform on HashiCorp-hosted VMs or self-hosted agents, Spacelift in containers on its public workers or your private workers, env zero on its hosted or self-hosted agents, Scalr on its shared runners or self-hosted agents, and Terrakube on its own executors.

Which of these tools are open source?

Stackorder, Atlantis and Terrakube are Apache-2.0. Stategraph Orchestration is MPL-2.0, and its self-hosted open-source build allows up to 3 active users a month per installation. OpenTaco is open core: MIT at the root, with a proprietary license for its ee/ directory. The Terramate CLI is MPL-2.0, and Terramate Cloud is a commercial SaaS. Terragrunt is MIT, and Gruntwork sells Terragrunt Scale. HCP Terraform, Spacelift, env zero and Scalr are proprietary platforms.

Which of these tools hold cloud credentials?

The Stackorder and Stategraph servers do not hold the credentials Terraform runs with: the CI runner does, and OpenTaco keeps them in your CI for its pull request automation too. Terramate and Terragrunt Scale leave them in your CI too: Terramate Cloud needs no access to your cloud accounts, and Terragrunt Scale uses OIDC credentials in CI. The Atlantis server holds them because it runs Terraform. HCP Terraform, and env zero and Scalr by default, store credentials or receive short-lived ones for each run; Spacelift assumes your role unless a private worker does it; Terrakube stores static credentials or issues its own OIDC tokens. Stackorder's shared outputs, off by default, give its server a read-only role on the state objects you list.

Try Stackorder on your own repositories

Free and open source under the Apache License 2.0. The getting started guide takes one repository from nothing to a first stackorder apply; the local demo runs on one machine with no GitHub App and no AWS account.