ChatInfra
LiveThe hosted SaaS core for running supported agent workspaces.
Hosted AI agent workspaces and guided setup
ChatInfra gives you a supported path to useful AI behavior: start with a focused workflow, run it in a hosted agent workspace, and connect implemented channels and tools with guided setup and safe defaults. Inspect, configure, connect, or take more direct control when your workflow needs it.
One platform, three lines
ChatInfra is one platform sold through three commercial lines
ChatInfra is live today. ChatInfra Edge and ChatInfra Sovereign are coming commercial lines that stay marked as roadmap until they are ready.
ChatInfra Edge
ComingThe on-prem appliance for running Agents and Probes on your premises.
ChatInfra Sovereign
ComingThe full-system self-owned license plus founder-led delivery for organizations with stricter control needs.
Included application and hosted Agent
Reach hosted ChatInfra from more than one surface
Included client application
ChatInfra Workstation
ChatInfra Workstation is a planned included first-party authenticated desktop client for existing hosted ChatInfra tenants. It is not yet available. Near-term, it brings messaging and alerts to the desktop; local-compute and agent-factory capabilities are future direction.
Included client application
ChatInfra Mobile
ChatInfra Mobile is an included first-party authenticated Android and iPhone client and reach app for existing hosted ChatInfra tenants.
First-party Agent
Calbot
Calbot is a first-party Agent inside hosted ChatInfra. This appointment-making AI concierge is live in production and shows what an on-rails workflow can do.
What runs
Agents do digital work. Probes report physical observations.
ChatInfra hosts two run types without making every visitor learn the full taxonomy first. A reusable class can become the specific instance that handles one workflow.
Agent
Agents do the work: they respond, schedule, draft, route, and call tools inside approved bounds.
Probe
Probes are physical monitoring or sensing endpoints that report observations into Agents for review and action.
Product pillars
The same four building blocks support each line
01
Hosted agent workspaces
A supported place to host, review, and improve the agent workflows you launch first.
02
Capability catalog
Outcome-driven, shareable capability pages that explain what each agent can do before setup starts.
03
Tool setup support
Guided help connecting calendars, contacts, and tools safely around the first workflow.
04
Messaging + connectors
Implemented WhatsApp messaging and Twilio voice connectors with gated availability, plus roadmap channels marked coming.
Twilio voice is implemented as a metered-add-on structure; its availability is gated, and production metering and charging are not claimed. Roadmap channels remain coming.
Desktop lineup
Desktop clients and desktop search serve distinct roles
Planned — not yet available
ChatInfra Workstation
The included desktop client for existing tenant members: near-term messaging and alerts, with local-compute and agent-factory capabilities as future direction.
Linux, macOS, and Windows live
ChatInfra Search Desktop
The desktop search application for shipped harness histories and local evaluator records.
Who it's for
- Operators and no-code builders who want reliable AI behavior
- Support teams who want faster responses with safe human handoff
- Teams planning WhatsApp messaging and Twilio voice connector setup
- Teams who need help configuring connectors and tools
What you can do first
- Launch one workflow for one outcome
- Plan setup for implemented XMPP, WhatsApp messaging, or Twilio voice capabilities while their availability remains gated
- Add tools (calendar/contacts) once the behavior is stable
You do not need to do everything at once. Start small, keep the first behavior reviewable, and expand once the dial setting is clear.
On-rails to autonomous
Start on rails and name the dial setting
ChatInfra does not need to pretend every workflow is fully autonomous. Calbot is the live worked example: an appointment-making AI concierge on rails in production. Its WhatsApp messaging and Twilio voice connectors are implemented separately, with gated availability pending named evidence gates.
Setup support keeps the first launch scoped, then the team can decide what to loosen only after behavior is stable.