About RelayOrb

A control plane for teams already operating production AI agents

Control plane for production AI agents. Versioned tool routing, schema validation, and audit logs for teams operating dozens of agents and tools.

What this does

RelayOrb sits between your agents and your tools. It gives you one place to validate schemas, route by capability version, enforce auth and policy, and inspect what actually happened when a call fails in production.

Who this is for

RelayOrb is for teams that already feel the pain of many agents calling many tools in production.

SRE / platform engineers running production AI agents

Your team is operating five or more internal agents that collectively call dozens of tools.

You need schema validation, versioning, and audit logs on every tool call. Ad-hoc wiring breaks at this scale. RelayOrb is the control plane.

Engineering leads accountable for agent reliability

You need to know which agent called which tool with which version when something fails.

You need to roll a tool version forward or back without redeploying every agent. RelayOrb gives you that pivot point.

Teams replacing in-house tool routers

You built your own gateway between agents and tools because nothing off-the-shelf existed.

Now it's tech debt. RelayOrb replaces it with versioned routing, OIDC-protected endpoints, and observable invocation state.

Who this isn't for

Solo developers and small teams. If you're wiring one agent to a handful of tools, ad-hoc wiring is correct. The control-plane overhead isn't worth it until you're managing many-to-many agent-to-tool relationships.
Anyone without production AI agents in flight. RelayOrb solves a problem you don't have yet. Come back when you do.
Hobbyists exploring agent frameworks. Start with the underlying frameworks (LangChain, LlamaIndex, plain MCP). RelayOrb sits one layer up; you'll feel its value only after the layer below is in active use.

How it works

Gateway validates caller auth, policy, and JSON schema before a tool call leaves the control plane.

Registry keeps track of healthy providers so routing decisions are versioned and observable instead of hard-coded.

Workers return validated responses, while RelayOrb records the invocation state you need for debugging, rollback, and audit.

Project status

RelayOrb is open source, deployed in production, and packaged for self-hosting through Terraform modules. The public site exposes operational evidence through the reliability report and public stats.

If you are already running production agents and need a control plane instead of more glue code, start with the Terraform module.