The secure platform for AI agents

Your AI agents,
under control.

Isolation, identity, access control and traceability, together in a single platform.

See the product Contact us
All your agentic workflows Dedicated servers anywhere, or self-hosted
From the creator of AutheliaOpen-source SSO & MFA · OpenID Certified™ · 29k+ GitHub stars
runyard · activity live
Isolated sandbox2 vCPU · 4 GiB · closed network
0 secrets in the agentinjected by the gateway
Agents are everywhere. Controls are not.

Four guarantees, one platform.

Where do your agents run?

Isolation

Each agent works in its own isolated space, on servers dedicated to you.

Who is acting when an agent acts?

Identity

Each agent has its own identity, which cannot be spoofed or borrowed.

What can they touch?

Access control

You decide, per agent type. Agents never hold your access keys.

What did they do?

Traceability

Every action is recorded, along with who did it.

The product

See it work.

Run agents from the browser, your own terminal, or your systems. Same guarantees everywhere.

app.runyard.ai
Fix bug #482 and open a PR.
› 6 steps · Thinking · Bash · GitHub

I reproduced the bug: the parser ignores time zones. Fix applied and tested.

  • Branch: fix/482-timezone
  • Tests: 128 passed, 0 failed
  • PR: acme/api#517 opened

GitHub access through the gateway: no token in the sandbox.

531 tokens · $0.07
Message the agent…Claude Opus 5.5 ▾
Claude Code · Opus 5.5 · /workspace
Runyard sandbox · closed network by default · attested identity
 
❯
In the browser, 24/7, or from your laptop, as usual.

Agents

AgentOwnerStateResourcesCost
Fix bug #482alice@acme.euidle● ready0% CPU · 474 MiB$0.07
Review PR #91bob@acme.eurunning● ready38% CPU · 1.2 GiB$0.22
Weekly sales reportcron · Mon 7amsleepingreleased2 CPU · 4Gi · 20Gi$0.29
Triage PagerDuty alertwebhookrunning● ready12% CPU · 640 MiB$0.04
Cost · 7 days$0
Tokens0
Requests0
Infra · now€0
InputOutputCache writeCache read
ChatWeb terminalYour terminalAPIWebhooksSchedules
Under the hood

The agent never sees your keys.

Every outgoing call goes through a gateway that verifies the agent's identity, enforces your policy, adds credentials outside the sandbox, and records everything.

Agentisolated sandbox
no secrets
attested identity
Runyard gatewaypolicy · secrets
audit log
token added
Your servicesGitHub, LLM, API,
MCP…
ActionGatewayOutcomeAgentTook
GET /v1/tunnels/previewtunnels200…7e92c3151 ms
PUT /v1/tunnels/previewtunnels201…7e92c31562 ms
POST /repos/acme/api/pullsgithub201…a41f09bd184 ms
GET https://evil.example/xegressdenied…a41f09bd0 ms
Customers

Who runs on Runyard.

specdriver What you meant to build, written down. Runs its agents on Runyard.
specdriver.io
Deployment

Dedicated servers or self-hosted.

Managed

Dedicated servers in Europe or worldwide, operated by us.

Self-hosted

On your own infrastructure. Nothing leaves your network.

Take back control of your AI agents.

Let's talk about your agentic workflows and how to secure them. Write to us:

contact@runyard.ai