CI Publishing without secrets
No long-lived credentials. CI proves identity at publish time via OIDC - nothing to rotate, nothing to leak.
Git is not a registry. You've always known that. Why does your Terraform/OpenTofu stack still treat it like one?
Terramantle gives authors insight, dependency management, and supply chain scanning they need to see what's consumed across your stack before they land in production environments with no complexity.
You're ready for Terramantle if you're managing 3+ modules across 1+ repos, want CI trust instead of long-lived secrets, and want signals your team actually sees.
No long-lived credentials. CI proves identity at publish time via OIDC - nothing to rotate, nothing to leak.
Review provider source code before it reaches your network. CVEs and malware caught before consumption.
Know what breaks before you deprecate. See every module consumer before making a change.
What you published is what gets consumed - forever. No silent drift from force-pushes.
No more credential sprawl across runners. Scoped tokens, RBAC, and audit logs that grow with you.
Keep your Github Actions or Gitlab CI workflows the same. No Terramantle agents or runners with hidden limits and costs
A state registry built for insight, discovery and safe day to day management.
Configure Terraform or OpenTofu's with standard HTTP backends. Your state is securely stored, versioned, and analysed - secret detection, public endpoint discovery run automatically on every push.
A simple click to force unlock a state file, no need to provide access to individuals to perform this action, keeping state files safe and secure.
Your infrastructure credentials pass through every provider you run. Think about that. Unvetted source code directly accessing your cloud environments!
Those providers are Go executables - compiled by strangers, distributed as Terraform plugins, pulling in dozens of transitive dependencies you've never seen or know about.
Mitigate unvetted source code reaching your production environments: scanning source code, going beyond standard CVE detection with AI models to inspect intent, and blocking malware before any binary is built and shipped to your environments.
AI-assisted analysis inspects provider source for execution intent, obfuscation, and known malware patterns beyond what signature scanning catches.
Most signed providers in the world only prove who signed - that a key signed some bytes. They say nothing about whether anyone looked at the code. A Terramantle signature is bound to a set of passed gates, so it makes a claim you can defend.
"Terramantle signed it" means "Terramantle compiled it, ran SAST + AI intent analysis + AV, and all passed - a defensible trust claim, not a blind one."
A signature normally means one thing: this key signed these bytes. It is an identity claim. It says nothing about whether the code was compiled cleanly, inspected, or scanned. The trust is blind - you are trusting the signer, not the contents.
Before the HSM signs, the signer independently asserts that every artefact hash maps to a gate verdict. The signature is bound to four passed gates - so it carries who signed and what was checked. The trust is defensible, not blind.
A gate computed inside the thing it gates is worthless, so the gates run trusted-side, independently of the build. Every gate has to pass before signing happens.
Semgrep runs against the full source tree with vendored rules. Static analysis before a single byte is compiled.
Claude inspects the whole source tree for backdoors, exfiltration, and obfuscation. Fail-closed: source that cannot be parsed, or a refused review, is treated as malware.
ClamAV scans the compiled binaries, re-hashed independently from object storage - not the source, the artefact you would actually run.
go.sum pinning is enforced and govulncheck runs reachability analysis, so a flagged CVE has to be reachable to count.
The builder compiles. It holds no signing key, has no LLM egress, and cannot reach the signer. A build RCE cannot forge a signature, because the thing that builds is architecturally cut off from the thing that signs.
A separate hasher computes the authoritative SHA256SUMS from object storage. The signer only signs hashes it can map back to a passed gate verdict.
Confirmed-malware-class gates - a ClamAV hit, or high-confidence malicious intent - need two admins to override. A single admin can only override low-severity findings.
We will not overstate this. The gates raise attacker cost - they are not proof of safety, and not authentication of upstream. The claim is precisely "compiled by us + passed our gates", not "this is safe". We think that honesty is the point. A signature that quietly implies "safe" is selling you something it cannot prove.
Run it yourself on Kubernetes or Docker when you need to own the infrastructure, enforce air-gap workflows, and operate disconnected registries for sensitive and regulated environments.
Highly Available by design, S3-backed object storage, so the service itself stays lightweight, easy to move, and simple to operate.
Open standards and cryptography bring providers and modules into regulated environments with proof they haven't been tampered with. No pointing engineers at public registries and trusting one person to vet what comes back.
Engineers keep using the same cli native tool behaviour and registry URLs they know, with no one-off wrapper to learn for self-hosted use.