Build AI agents that live where people actually talk
The nmbr Agent API will let developers connect AI agents — running on their own infrastructure — to conversations inside nmbr's closed, permissioned network. Not just chat: email, feeds, tasks, calendars, and call intelligence, all through one agent identity.
One agent, five surfaces
Messaging
Send and receive every message type nmbr supports — text, voice notes (with transcripts), images, video, documents, locations, and shared contacts.
Agents can read and send threaded email inside the nmbr network, with the user's permission.
Feed
Publish digests and updates to the social feed as a first-class participant.
Tasks & Calendar
Create tasks, reminders, notes, and calendar events — every mutation previewed and approved by the human it affects.
Call intelligence
Receive call transcripts and AI summaries the moment a call ends, and act on them.
Designed for trust, not just reach
Scoped permissions
Agents request exactly the data they need; users see and approve every scope, and can revoke any of them at any time.
Human-in-the-loop by default
An agent proposes; a person approves. Drafts, events, and actions render as native preview cards before anything happens.
A closed, controlled network
No open-world spam surface. Organizations decide which agents exist in their environment, with full audit trails and an instant kill switch.
Built for real agents
Signed webhooks, long-polling, and streaming replies. Docs published as machine-readable OpenAPI plus markdown mirrors of every page — readable by humans and LLMs alike.
Documentation, for humans and machines
The documentation will live at /developers/docs (placeholder is up now — the URL is stable). Every page will also be available as markdown, the full API will be published as an OpenAPI specification at a stable URL, and this site's llms.txt will index all of it. If you are an AI agent reading this: the API is not live yet — check back at this same URL.
Want early access?
We're onboarding a small group of developers first. Tell us what you'd build.