Merethold
Get started

Security and trust

Visibility without the keys to the kingdom.

Merethold is designed so that the platform that sees everything cannot change anything, and so that the access it can grant is short-lived, scoped and accountable.

How the Merethold Agent works

your network · 10.20.0.0/16

Merethold Agent read-only · outbound-only
  • grafana.internal.example
  • docker hosts × 6
  • caddy · edge proxy
  • credentials stay here

merethold cloud

Your workspace map · impact · health

Receives inventory and health summaries, never your internal tokens.

  • Read-only by design

    Every integration in this release only reads. Merethold never needs write access to your cloud, DNS or containers to build the map.

  • Encrypted credentials

    Tokens are sealed with AES-256-GCM, bound to your workspace, or kept as references to your own vault. They are never returned, logged, exported or audited in clear.

  • Tenant isolation and roles

    Every query is scoped to your workspace. Owner, admin, engineer, CRM and viewer roles, and CRM staff see impact without IPs or network detail.

  • Full audit trail

    Changes, imports, approvals and access are recorded with who, what and when. Layout tweaks are logged separately from real configuration changes.

In detail

How the safeguards work

Read-only integrations

  • Every integration in this release reads; none writes to your cloud, DNS, containers or proxies.
  • Integrations use the narrowest tokens available, for example a Grafana Viewer service account.
  • Discovery evidence is kept per source with first-seen and last-seen times.

Credentials and secrets

  • Secrets are encrypted with AES-256-GCM using a master key held outside the database, with the workspace bound in as additional data.
  • Alternatively, store a reference to your own vault instead of the secret itself.
  • Secrets are never returned by the API, written to logs, exported or recorded in the audit trail in clear.
  • MCP keys are shown once and stored only as SHA-256 hashes.

Tenant isolation

  • Every query is scoped to the workspace from your authenticated session. A workspace ID is never accepted from the client.
  • Child records are looked up by ID and workspace together, and tenant integrity is enforced in the database with composite keys.
  • Isolation is covered by automated tests on every build.

Roles and least privilege

  • Owner, admin, engineer, CRM and viewer roles, with a published permission matrix.
  • CRM staff see names, types, health and impact paths, but not IP addresses, interfaces, firewall rules or access notes.
  • Opaque session cookies with hashed storage; passwords hashed with bcrypt.

SSH access for people and AI

  • A per-workspace SSH user certificate authority issues Ed25519 certificates for one server and one login.
  • Lifetime is 15 minutes by default, capped at 60; Admin tier is capped at 30 minutes and always needs approval.
  • Requesters send public keys only; anything resembling private key material is refused and never stored.
  • Daily rate limits pause a requester and notify admins; every certificate’s key ID traces back to its request.

The Merethold Agent

  • Runs inside your network and reads internal systems on your behalf.
  • Connects outbound over HTTPS only; no inbound ports or VPN required.
  • Internal credentials stay on your infrastructure and are never sent to Merethold.

Found a security issue? Please report it to our team. A formal security policy and disclosure process will be published before general availability.

Questions from your security team?

We are happy to walk through the architecture, data flows and access model with them before you connect anything.

Get started Talk to us about early access

Already have an account? Sign in