Open Bee AI

Getting started

Sign up at openbee.ai/sign-up and configure an AI provider key (or BYOK) in Settings → BYOK.
Build an agent in /w/<your-slug>/agents — write a prompt template with {{variables}}, pick a model + thinking level, test it, then publish. (Author reusable prompts first in Prompt Studio.)
Call it from your systems: create an API key in Settings → API clients, then POST /api/v1/agents/<id>/run with a Bearer token and your inputs — or wire it into n8n.

Choose your path

Self-hosting guide →

Run the open-source agent daemon on your own machine — your code, your keys, your hardware.

API reference →

Public REST API to run your agents (POST /api/v1/agents/:id/run) + the daemon protocol. Bearer-token auth.

Source code ↗

The whole monorepo on GitHub. Daemon is Apache-2.0; cloud surface is source-available.

Security model →

Encryption, isolation, sub-processors. What we protect and how.

Concepts

Workspace. Your billing and isolation boundary. Members, prompts, agents, runs, and API keys all scope here.
Prompt. A reusable, versioned template authored in Prompt Studio with {{variable}} placeholders. Reuse it inside an agent.
Agent. One or more ordered steps. Each step has its own instructions, prompt template, model, and reasoning (thinking) level; a later step can reference {{previous}} or {{stepN}}.
Public API. Call any published agent with an API client key: POST /api/v1/agents/<id>/run. Returns the output and usage as JSON; rate-limited per key.
n8n. Two-way automation — n8n calls your agents via the public API, and agents can POST their result to an n8n webhook on completion.
BYOK. Bring your own Anthropic / OpenAI / Google key in Settings → BYOK; agent + prompt runs route through your provider account, never our shared quota.
Daemon. An optional Apache-2.0 process you run via open-bee start to execute tasks on your own hardware — code and files never leave your machine.

Quick CLI reference

# Install (from a checkout of github.com/openbee/openbee)
pnpm install
pnpm --filter @openbee/agent-runtime build

# Link to your workspace
pnpm --filter @openbee/agent-runtime open-bee link \
  --workspace your-slug \
  --key obee_your_api_key

# Run the daemon
pnpm --filter @openbee/agent-runtime open-bee start

# Inspect status
pnpm --filter @openbee/agent-runtime open-bee status

Where to next

Self-hosting covers prerequisites, environment setup, and operating the daemon at scale.
API reference documents every endpoint including request/response shapes and rate-limit headers.
Changelog tracks every release.
Stuck? Email hello@openbee.ai or file an issue on GitHub.