Skip to content

Product · In use

Build, organize and trigger AI-powered automations end to end, without writing code.

CAL lets teams compose automations from AI agents, visual drag-and-drop workflows, Custom APIs and external API connections, then run them from the screen, on a schedule, or from outside systems over REST and webhooks. It works across OpenAI, Anthropic, AWS Bedrock, Google Gemini and Azure, and runs as a hosted cloud service or self-hosted on Docker.

Request a 20-minute scoping call

What you get

Agents on any provider
AI agents with a prompt, model, tools and input and output schemas; swap any model.
Visual workflows
Drag-and-drop workflows with branching, loops, API calls and sub-workflows.
Custom APIs and external connections
Typed backends of your own, used by workflows and agents, and connections to outside APIs.
Triggers
Start a run from REST, a chat widget, a webhook or a schedule.
Templates
Portable bundles of Custom APIs, external APIs, agents and workflows.
Your own provider keys
Each organization can bring its own AI provider keys.
Single sign-on
Okta and Microsoft Entra ID.

How it runs

Cloud: runs on AWS in us-east-1, on one EC2 server with its database on Amazon RDS PostgreSQL; many organizations share it, each organization is the tenant, the application keeps each one’s data to itself, and only CAL staff can see across organizations. Self-hosted: runs in the customer’s own Docker install with PostgreSQL, one customer per install. Provider keys and connection secrets are stored encrypted. How a customer’s data comes back at the end of a contract is not yet stated.

What it does not do

  • No automatic failover between AI providers and no cost-tier routing.
  • One cloud region; no multi-region cloud.
  • No customer-managed encryption keys ("bring your own key" means AI provider keys).
  • No free tier.
  • Two-factor sign-in can be set up but is not enforced at login.

Talk to us →

All products

Start with a 20-minute scoping call.

Tell us what people need to learn, who needs it and when it must be ready.

Request a 20-minute scoping callCapability statement