BYOA · bring your own agent

Your agent.
We run the rest.

Already built a support agent — Claude Code, Codex, your own stack? Connect it to bot.dev and it answers real customers, while we handle delivery, monitoring, guardrails and human hand-off. You bring the brain.

Your customer chats in our widget; we route and watch it in our hub; the question reaches your own AI through one outbound CLI tunnel — and the reply comes straight back.

1

command to put your agent in front of customers.

No inbound firewall changes, no rebuild. The connector dials out from your infra, receives each customer message, and posts your agent's reply back into a real chat widget.

bash — your infrastructure
# point bot.dev at whatever answers your questions $ bot connector \ --agent-id acme \ --backend-cmd "claude -p" subscribed · answering ch-v-71ad95e8 live reply delivered to widget (0.8s)

You bring the answer. We bring everything around it.

The unglamorous production layer you'd otherwise build and run yourself — done, from day one.

The chat surface & live delivery.

A real embeddable widget and the real-time transport to the customer's browser — the part that's fiddly to get right and easy to get wrong.

Your backend Claude Code Codex Your own API …anything on a CLI

Live monitoring.

Every conversation your agent is handling, visible in one shared inbox — sentiment and confidence flagged.

Human hand-off.

When your agent should step aside, it escalates to your team — with the topics you choose always routed to a person.

Guardrails & audit.

Full transcripts of every exchange, so when your agent answers you can see exactly what it said and when.

Same binary, hosted or self-run.

Run the connector yourself for full control, or let us host it. Want us to bring the AI too? That's Managed mode.

Bring your agent. Keep control.

Have an invite code? Create your workspace and connect your agent — live in seconds. Otherwise, ask us for one.