Home How it works Connect Developers Contact
Setup
⚡Keep your AgentBee plugged in until you see “Done”. Don't unplug during setup — same as a phone during an update, pulling power can damage it.
1

Connect your AgentBee

Plug it into USB. Use Chrome or Edge (Web Serial). Nothing is sent anywhere — setup runs entirely in your browser.

2

Genuine check

Your browser verifies the device is real and untampered — its attestation must chain to the AgentBee Root, held in our HSM. Then it challenges the device to sign a fresh random number live.

3

Create your key (on the device)

The private key is generated and locked inside the chip. It never leaves the device and never reaches us. Only the public key comes out.

4

Connect it to Claude

One command. After this, your AgentBee guards every critical AI action — for life.

claude mcp add agentbee -- npx -y @agentbee/mcp
ACopy the command above.
BPaste it into your terminal (or Claude Code) and press Enter.
CDone. Now when your AI tries something critical — delete a database, push code, move money — your bee lights up and asks. Flip APPROVE to sign it, or DENY to block it. Two switches, your call.
Claude Desktop? Add the same line under mcpServers in claude_desktop_config.json.
Want it to gate actions automatically? Install the PreToolUse hook.  ·  Developers: npm i @agentbee/sdk · docs.
Your private key never leaves the device. We store no keys, no device data, and no email. There is nothing about you for us to lose.