BMA (BeeOS Managed Agents) is an open-source agent runtime. Mix any brain with any hands:
  • Brain: any model and any agent loop.
  • Hands: cloud sandboxes, Linux, Windows and macOS cloud computers, or your own machine.
Your agent can use a computer, you can watch it and take over through a remote desktop (VNC or WebRTC) or a terminal, the scheduler is replaceable, and every step is recorded in a durable event log that you can read or stream.

What you get

Brain, hands, session

BMA keeps three things separate: Because the session lives outside the machine, a machine can be paused, reclaimed, rebuilt or restored without losing the conversation, and the brain can be replaced without changing how clients read events. You define an agent (model, instructions, tools), start a session of it in an environment, and read the session’s events as it works. Your app talks to the Managed Agents API under /v1 with the official Anthropic SDK (Python, TypeScript) or plain HTTP. A second protocol, the Executor Protocol, connects machines and external brains to BMA; you only need it when you build your own hand or brain, using bma-executor-sdk.

Base URLs and keys

The hosted service currently runs as staging at https://api.staging.bma.beeos.ai. The BMA team gives you an organization API key (bma_org_v1_…) and a project. In these docs:
The official SDKs default to their own vendor’s API host, and bma-executor-sdk to https://api.bma.beeos.ai, which is not live yet. Always pass the staging base URL explicitly, https://api.staging.bma.beeos.ai, as base_url / baseURL (and to bma-executor-sdk); each SDK adds its own path. Send x-bma-project-id when your key sees more than one project.

Start here

Quickstart

Create an agent, start a session, stream its reply and send a follow-up. Python, TypeScript and curl.

SDKs

Which SDK to use for what: the official Anthropic SDKs and bma-executor-sdk

API reference

The /v1 API: base URL, headers, errors and every BMA extension endpoint

Executor & brains

Bring your own machine or your own agent loop