Metadata-only by default
Guard findings carry hashes of file paths and never file contents. Prompt text is inspected in memory and never stored in a finding.
Trust
Security tools should earn trust, not assume it.
How Runeward Labs handles your data, how it proves what happened, and how to report a problem.
Data handling
Collect the least, prove the most.
Digests, not images
Screenshot correlation computes a digest locally and discards the image bytes.
Signed, verifiable evidence
Rune Box audit events are hash-chained and Ed25519-signed; Guard devices upload signed, chained findings.
Tenant-scoped access
Server-enforced roles and tenant isolation in Guard Cloud and Rune Box; keys are never stored in the browser.
Signed supply chain
Signed detection catalogs for Guard and a signed, checksum-verified installer for Rune Box.
Your infrastructure
Self-hosted deployments keep evidence inside your environment. This website sets no cookies and loads no third-party resources.
Privacy statement
What Gurd collects, and where it is processed.
Rune Guard monitors AI agents, not developers. By default it collects metadata about what agents do and never reads a person's chats or prompts. What people share with AI tools, and the content of agent sessions, are collected only when the organization turns on the matching opt-in.
- Hosted Guard Cloud
- Processed in the EU.
- Self-hosted Guard Cloud
- Processed wherever you run it.
- In detail
- What each collector gathers and what a device sends.
Software bill of materials
Every Rune Guard release ships signed CycloneDX SBOMs. See what each one lists and how to verify it.
Trust centre
What is true today.
Where each part of the product stands today.
| Topic | Today | Status |
|---|---|---|
| Rune Box SBOM | Every Rune Box release archive ships an SBOM, alongside cosign-signed checksums. | Live |
| Guard SBOM | Each Guard release publishes CycloneDX SBOMs for the agent, the Cloud crates, the Cloud image and the edge image, each signed with the release key. The newest release’s are linked from the trust centre. | Live |
| Vulnerability disclosure | Report privately, through GitHub private vulnerability reporting or to the maintainers. Never put live secrets, screenshots or exploit details in a public issue. | Live |
| Privacy tiers | Metadata about what agents do is always on. What people share with AI tools, and redacted excerpts of agent session content, are off by default, each behind its own per-company opt-in in the signed catalog. | |
| Where data is processed | Hosted Guard Cloud processes telemetry in the EU. A self-hosted Guard Cloud processes it in the region you run it in. | |
| Retention on the device | The local findings log keeps 30 days in a hash chain. The screenshot monitor keeps its audit log for 1 to 365 days, as you choose. | |
| Retention in Guard Cloud | On Enterprise, Guard Cloud keeps findings, session steps and health events for replay and audit, with no automatic deletion. On Startup they are deleted after 90 days (an administrator can choose 30 to 90), except findings linked to an incident, which stay while the incident exists. Daily device counts are deleted after 200 days by default. Opt-in session content is kept only for its opt-in’s or incident’s retention period. Self-hosted operators control their own PostgreSQL. |
Responsible disclosure
Found a vulnerability?
Please report it privately rather than in a public issue. Follow the Rune Box security policy for reporting instructions; for Guard, email [email protected].
Compliance status
Runeward Labs holds no security certifications yet. Framework tags on findings are Runeward Labs’ own, to help triage.