Menu
Blog How it works Use Cases
agentsh
Open-source runtime for AI agent security
Beacon
AI endpoint visibility and control
Watchtower
Enterprise control plane for agentsh and Beacon
Request Access

Execution-Layer Security
for AI workloads.

Control what AI can access, run, and connect to — at runtime.
Supervised copilots on endpoints. Unsupervised agents everywhere else.

One control plane. Local enforcement.

  • Control what matters Allow, prompt, block, or redirect — by destination, command, tool, or workload.
  • See everything Every AI-triggered connection, command, and file change, fleet-wide.
  • Prove it happened Audit trails, SIEM export, and an emergency brake when you need it.
  • Steer, don't just block Redirect to approved registries and endpoints so work keeps moving.
runtime decisions
Cursor endpoint registry.npmjs.org Prompted
Claude endpoint unknown-mcp.io Blocked
Agent pipeline exec terraform apply Prompted
Manus container connect api.github.com Allowed
Langgraph dev env connect pastebin.com Blocked
Codex sandbox internal-registry.company Redirected
Two contexts

Two execution contexts. Same runtime control.

Supervised

Supervised AI on endpoints

Beacon

Copilots and desktop tools run with employee credentials.
Beacon adds guardrails and approvals to keep endpoints productive and safe.

Learn about Beacon
Unsupervised

Unsupervised agents anywhere they run

AgentSH

Headless agents execute fast with no user to prompt.
AgentSH enforces least privilege at the syscall level — in CI, containers, pipelines, and dev environments.

Learn about AgentSH

Watchtower governs both with centralized policy, kill switch, and SIEM forwarding.

Learn about Watchtower →
See all use cases →
Runtime decisions

Runtime decisions in plain English.

Cursor endpoint registry.npmjs.org Prompted
Claude Desktop endpoint unknown-mcp.io Blocked
Agent pipeline exec terraform apply Prompted
Manus container connect api.github.com Allowed
Langgraph dev env connect pastebin.com Blocked
Cursor endpoint internal-registry.company Redirected
Why now

AI already executes inside your company.

It's on endpoints

Developers installed Claude, Cursor, ChatGPT. IT did not provision it. Security cannot see it.

It's moving into automation

Agents are being added to CI, pipelines, ops, and internal workflows. No UI. No supervision. Full blast radius.

Blocking does not work

Users route around walls. You need guardrails that steer toward approved workflows and enforce least privilege.

FAQ

A few quick answers.

What's the difference between Beacon and AgentSH?

Beacon secures supervised AI on endpoints. AgentSH secures unsupervised agents wherever they run — CI, containers, pipelines, and dev environments.

What does Watchtower do?

Watchtower is the command center. It distributes policy, routes approvals, exports to SIEM, and provides a fleet-wide kill switch. Beacon and AgentSH enforce locally at execution time.

What do you mean by "steer"?

Steering redirects AI to approved alternatives. It keeps users productive and prevents retry loops that happen when agents keep hitting a hard block.

More questions →

Get on the early access list.

Tell us your environment — endpoints, CI, which AI tools, which agents. We'll prioritize the right fits.

Prefer email? hello@canyonroad.ai
No spam. One email when it's real.