@sente-labs/cli

The Sente CLI -- give your AI agents managed email identities (AgentAccount). Create an email address your agent owns on sente.run, then send, receive, and react to mail -- verification links, OTPs, replies -- from your terminal, scripts, or Claude Code.

Install

npm install -g @sente-labs/cli

Requires Node.js 18+.

Building an agent that should set this up itself? Point it at the onboarding skill: https://sente.run/skill.md.

Quickstart

# 1. Sign in (opens the browser, stores an API key in ~/.sente/credentials)
sente login
sente whoami

# 2. Give your agent an identity -- a real, reusable email address
sente identity create --name "Support Bot" --local-part support-bot
# -> [email protected]

# 3. Read its mail
sente inbox --identity support-bot                 # recent messages
sente wait --identity support-bot --timeout 60     # block until the next message
sente listen --identity support-bot                # stream messages as they arrive

# Verification emails: --otp / --magic-link block for the verification message
# and print JUST the extracted code / link (pipeable):
CODE=$(sente wait --identity support-bot --otp --timeout 60)

# 4. Send from it
sente send --identity support-bot --to [email protected] \
  --subject "Hello" --text "It works."
# Reply in-thread to an inbound message:
sente send --identity support-bot --to [email protected] \
  --subject "Re: Hello" --text "Following up." --reply-to <messageId>

Use --json on any command for machine-readable output.

Webhooks (push delivery)

Register your HTTPS endpoint and Sente POSTs events — message.received for each inbound email, and run.blocked when a registration run needs a human (see below). Verify the X-Sente-Secret header on every delivery.

sente webhook register --url https://my-app.com/hooks/sente --identity support-bot
sente webhook list
sente webhook delete <id>

Developing the handler locally? Relay live inbound mail to it with no public URL -- listen --forward POSTs each message in the exact webhook envelope (with x-sente-secret taken from SENTE_WEBHOOK_SECRET, if set):

sente listen --identity support-bot --forward http://localhost:3000/hooks/sente

Or run a shell command per message (message JSON on stdin):

sente listen --identity support-bot --exec './handle-message.sh'

Register & sign in to apps

Have Sente drive a browser to sign your agent up at a third-party app under the identity's email, completing the email verification itself:

sente register https://app.example.com --identity support-bot   # → a durable account
sente registrations --identity support-bot                      # list accounts + latest run
sente relogin <registrationId>                                  # re-authenticate later
sente credentials <registrationId>                              # print the vaulted login

sente register blocks until the run finishes. Add --confirm-before-submit to keep a human in the loop for the final submit — the agent fills the whole form, then the run pauses blocked with a live-view URL for you to click submit yourself (you accept the app's terms), and sente run resume <runId> continues.

To be notified when any run needs you (confirm-before-submit, CAPTCHA, …) without standing up a webhook server, run the watcher — it desktop-notifies (macOS Notification Center; terminal bell elsewhere) with the live-view URL and resume command:

sente watch                       # all identities
sente watch --identity <ref>      # one identity
sente watch --interval 10         # poll every 10s (default 5)

Prefer push? Subscribe a webhook to run.blocked (payload carries the block code, liveViewUrl, a dashboard actionUrl, and holdExpiresAt):

sente webhook register --url https://my-app.com/hooks/sente --events run.blocked

Auth

sente login stores the API key in ~/.sente/credentials (mode 0600). The key authenticates as your whole organization -- don't paste it into chats or commit it. To hand it to your app without echoing it:

echo "SENTE_API_TOKEN=$(sente token)" >> .env

For CI or non-interactive use, set env vars instead of logging in:

export SENTE_API_TOKEN=sk_sente_...
export SENTE_GATEWAY_URL=https://api.sente.run   # optional override

Commands

| Command | What it does | | --- | --- | | sente login | Browser sign-in; mints and stores an API key | | sente whoami | Show the logged-in account + organization | | sente token | Print the stored API key (for piping into .env) | | sente identity create\|list\|show | Manage email identities | | sente inbox --identity <ref> | List recent messages | | sente wait --identity <ref> | Block until the next message (long-poll); --otp / --magic-link print just the code / link | | sente listen --identity <ref> | Stream messages; --exec / --forward to react | | sente send --identity <ref> | Send email (--reply-to threads replies) | | sente webhook register\|list\|delete | Manage push delivery to your endpoint | | sente register <appUrl> --identity <ref> | Sign up at a third-party app under the identity (--confirm-before-submit keeps a human on the final submit) | | sente registrations | List accounts and their latest run status | | sente relogin <registrationId> | Re-authenticate an existing account | | sente credentials <registrationId> | Print an account's vaulted login | | sente run show\|resume\|abort <runId> | Inspect and control a register/login run | | sente watch | Desktop-notify when a run needs you (blocked); --identity, --interval |

<ref> is an identity id, email, or local-part.

Documentation

• Home: https://sente.run
• Agent onboarding skill: https://sente.run/skill.md
• Issues: https://github.com/sente-labs/agentic-browser/issues

License

MIT