Runewrd Labs
ConsoleGet early access
MenuClose

Platform

Agent security across the whole lifecycle.

Rune Guard secures agents where people use them: on laptops, in IDEs and in browsers. Rune Box secures agents where they run work, in governed, isolated environments. Both record evidence the same way.

Rune GuardRune Box

Discover · Assess · Govern · Enforce · Investigate · Improve

What each product contributes at every stage.

What Guard and Rune Box each do at every stage, from discovery to improvement.

Discover

Find the AI tools, harnesses and MCP servers on your endpoints, sanctioned or not.

Guard
Catalog-driven inventory, behavior-based discovery of unlisted tools, MCP server inventory.
Rune Box
Every governed run is attributed to tenant, actor, parent run, provider and model.

Assess

See risky configuration before it becomes an incident.

Guard
guard scan reports bypass and never-ask modes, disabled sandboxes, broad allow rules and hidden instructions.
Rune Box
runeward doctor checks policy, runtime, image, state and secret sources before launch.
Live

Govern

Decide what agents may do, with humans in the loop where it matters.

Guard
Pin approved MCP tool definitions; one-shot operator approvals for enforcing hooks.
Rune Box
Charters with glob, CEL, OPA/Rego or signed OCI policy bundles; allow, deny or require approval.
Live

Enforce

Stop the risky action and let the developer keep working.

Guard
Harness hooks deny credential reads, and block prompts that carry secrets when the human-to-AI sharing opt-in is on, where the harness supports it.
Rune Box
Isolated sandboxes, deny-by-default egress, and wall-clock, exec, token, cost and retry budgets.

Investigate

Follow the evidence with source, timing and outcome stated plainly.

Guard
Findings with their reported sources, hashed paths and device timelines in the Guard Console.
Rune Box
Append-only, hash-chained, Ed25519-signed audit events and portable evidence exports.

Improve

Turn real evidence into tighter policy.

Guard
Signed detection catalogs delivered to enrolled devices.
Rune Box
policy learn proposes exact rules from verified evidence for a human to review.

Architecture

Two products. Each works on its own.

Adopt Rune Guard or Rune Box without the other. Use both and they cover agents from the laptop to the execution environment.

Rune Gurd

Secures agents where people use them: laptops, IDEs and browsers across your organization.

  1. Gurd agent

    Rust endpoint agent: inventory, sensor, harness hooks, MCP checks and the local scan.

    Runs on
    • macOS
    • Linux
    • Windows

  2. Gurd Cloud

    Self-hostable backend: device enrollment, status, findings and signed catalogs.

    Runs onYour infrastructure

  3. Gurd Console

    Live

    Device health, findings and evidence timelines for security teams.

    Runs onAny modern browser

  4. Managed Gurd Cloud

    Guard Cloud, run for you by Runeward Labs.

    Runs onHosted by Runeward Labs
Explore Rune Guard →

Rune Box

Secures agents where they execute work: governed, isolated environments for any agent or framework.

  1. CLI and control plane

    Live

    Policy, approvals and budgets for every tool call. Binds to loopback by default, with RBAC and OIDC for teams.

    Runs on
    • macOS
    • Linux

  2. SDKs and MCP server

    Live

    Python and TypeScript SDKs, framework adapters, and governed tools for any MCP-capable IDE or agent.

    Runs onYour agent’s runtime

  3. Isolated environments

    Live

    One Citadel per agent or worker, with its own mounts, network rules and limits.

    Runs on
    • Docker
    • OrbStack
    • Podman
    • Kubernetes

  4. Signed audit

    Live

    Append-only, hash-chained, Ed25519-signed events you can export and verify anywhere.

    Runs onPortable files
Explore Rune Box →

Better together

Guard covers the agents people run on their own devices; Rune Box covers the agents that run work on their own. Use both and nothing falls between them, and a finding means the same thing in either product.

Honest evidence

Words that mean exactly one thing.

Security teams lose trust in tools that call every warning a prevented attack. Runeward Labs keeps its evidence vocabulary strict across both products.

Observed
The sensor or a hook saw it happen. No claim about intent or harm.
Decision
Policy returned allow, warn, deny or require-approval. A block decision is not proof the action was prevented.
Blocked (reported)
The enforcing component reported that the action did not run.
Evidence basis
The reported source and recorded method used to link an event to an AI tool. It does not establish intent, harm or causality.

Your agents already have access. Ward off the risks.

Join Guard early access for your team, or run the free local scan today. Agents in automation too? Add the open-source Rune Box.

Get Guard early accessRun the free scan