Orb

How it works

You have data in different places: APIs you use, services you built. Orb puts them behind one endpoint. Your agent gets one address and one key, and can use what you let it.

1. Add a data source

Declare an HTTP API in Orb's service config: a name and where to reach it.

2. Connect PocketBase

Orb uses PocketBase to store access keys and check every incoming request.

3. Create an API key

Make a key and choose which services it can read. Keys are rate-limited, and requests are logged.

4. Give your agent the endpoint and the key

Each API you document gets an info page with a copy button. It copies a short brief (address, auth, endpoints, example request) to paste to your agent. The key itself goes through your agent's own credential setup, not chat.

How Orb fits between your APIs and your agentYour APIs on the left point to Orb in the middle. Orb points to your agent on the right.Your APIsOrbone endpoint, one keyYour agentHow Orb fits between your APIs and your agentYour APIs at the top point down to Orb in the middle. Orb points down to your agent at the bottom.Your APIsOrbone endpoint, one keyYour agent

What Orb is not

  • Not a hosted service. You run it yourself.
  • Not an agent. It gives your agent access; it doesn't act on its own.
  • Not the only option. Projects like IBM ContextForge and MCPJungle solve related problems. Orb is a personal setup, built around one person's data and agents.

Where it stands

Orb is a personal project in active use. This page describes how it is meant to work; the guides say what to run. If something differs, the code is the reference.