Skip to main content

Organization — Summary

The Organization is the administration UI for your Keen sandbox. It lives at https://<your-domain>/org. From here you manage who can log in, what they can do, where flows live, which agents exist, who can talk to them, and how logging behaves.

This page is a one-screen overview. Each section below has its own dedicated page — click through for the detail.


Two environments

EnvironmentWhat it's forNotes
Sandbox (development)Building and debugging flowsLimited resources. Debugging enabled — JS code is deployed and can be inspected via the Keen Builder VS Code extension.
ProductionRunning flows for real usersMicroservice architecture with horizontal scaling. No debugging. Same shape as sandbox, just no debugger.

JS is the main language for everything you write. Flows and scripts are authored in the Keen Extension and deployed to your sandbox via a deploy command.


Pages in /org

The sidebar groups the admin surface into seven areas. Each has a list page and a Create link.

PageWhat it managesDetail
AccountsUsers who log into /orgaccounts.md
RolesPermissions assigned to accountsroles.md
AI SpacesContainers for agents — one space = one VS Code projectai-spaces.md
AgentsThe entities users actually talk to in the Chatagents.md
ConsumersChat-only users (no /org access)consumers.md
Logic Fields(not yet documented — placeholder)logic-fields.md
LogsWhat the runtime captures from system + flow executionlogs.md
SystemPlatform-level settings (Consumer Settings, Google Connect)system.md

Reference screenshots of every page are in guides/system-information-snapshots/ (file dates 2026-04-21).


Mental model — how it all fits

The Organization is a stack of contracts the runtime executes against:

  1. Accounts + Roles define who can edit /org.
  2. AI Spaces define the boundaries within which flows are stored and executed.
  3. Agents declare an entry into a specific Space's flow code.
  4. Consumers declare who can talk to which Spaces (and therefore which Agents).
  5. Logs declare what the runtime persists from system + flow execution.

Every time a Consumer sends a message to an Agent in the Chat:

  • Their Consumer record lets them through to the Space.
  • The Agent record tells the engine which Start Flow / Start Node to enter.
  • The flow's nodes execute, writing into the dictionary, talking to LLMs, calling system libraries.
  • Logs capture whatever's Engaged.
  • The response goes back to the Chat.

The flow developer's job is to design the graph. The Organization is what makes that graph reachable.


Common gotchas

  • Agent ID must be lowercase AND must exactly match the agentName in the Agent Node inside its Start Flow. Mismatch → "Invalid agent data!" from the Chat. (details)
  • Role / permission changes log the user out immediately — they have to log back in. (details)
  • There can only be ONE Owner. Administrators can be created only by the Owner. Administrators cannot create more Administrators. (details)
  • Error logs cannot be disabled. That row is always on. (details)
  • An Agent belongs to exactly one Space. A Space can have many Agents. (details, agents)