Skip to main content

Squads

The Squads group is the collaboration hub of the sidebar: squad rosters, the agent registry, your personal agents, and run history. Sidebar items are role-filtered — Squads, Org Agents and Sessions require the agent admin role or above, while My Agents and My Sessions are visible to every signed-in user.

Squads (/squads)​

/squads lists every squad — groups of agents that share memory, conversation context, and artifacts, and can delegate tasks to each other. Requires agent admin or above.

  • Table view — squad name and description, scope (Org-wide or Personal), member avatars with the squad lead highlighted, member count, last updated, and an Active/Archived status badge. Clicking a row opens the squad's detail page.
  • Filters — a segmented All / Org-wide / Personal switch, plus a Show archived toggle. A count of matching squads sits on the right.
  • New squad — an inline creator with name, an optional one-line description, and scope (Org-wide or Personal). A Let AI suggest from a one-line brief option fills the name and description from a short prompt.
  • Browse marketplace — opens a drawer of installable squad templates. See Marketplace.

Org Agents (/agents)​

/agents is the organisation-wide agent registry: every agent in the workspace. Requires agent admin or above. For agent concepts and configuration, see Managing agents.

  • Table view — agent name and short ID, team, model (provider and model, or Default), bound channels, last updated, and status (Active, Inactive, or System for platform-managed agents, which sort first). Clicking a row opens the agent's detail page to inspect, edit identity, run, or view logs.
  • Filters — search by name or template, a segmented All / Active / System switch, and a team dropdown when teams exist.
  • New agent — a modal that starts from a template (pre-filling identity, model settings, and skills) or blank, with name, team, identity in Markdown (with AI assist from a one-line brief), and model provider/model. The identity field shows a live character counter against the organisation's identity limit — text beyond the limit is truncated at run time, and a configure limit link jumps to Settings.
  • Row actions — Edit, View logs, Enable/Disable, and Save as template (name, description, and category — the template then appears in the create modal). Delete is offered only for inactive, non-system agents and requires typing the agent's exact name; it permanently removes the agent with its tools, config history, schedules and assignments.
  • Browse marketplace — installs agent templates from the configured marketplace.

My Agents (/my-agents)​

/my-agents lists your personal agents — agents that use your own connectors and have isolated memory, visible and usable only by you. Available to every signed-in user.

  • Table view — agent, model, bound channels, last updated, and status, with a search box and a count.
  • New agent — create from a template, clone from an org agent (your personal copy gets isolated memory and uses your own connectors, with identity and model pre-filled), or start blank with a name, identity, and model.
  • Row actions — Edit, View logs, Enable/Disable, and Delete for inactive agents (type-the-name confirmation, irreversible).
  • Browse marketplace — templates installed from here land as personal agents on this page, not on the org-wide registry.

Sessions (/sessions)​

/sessions is the org-wide session log — every agent run across the organisation. Requires agent admin or above.

  • Filters — status (Active, Completed, Error) and channel (Slack, Teams, email, and other connected channels).
  • Table view — session ID, a type badge (User, Scheduled, Proactive, or Heartbeat), channel, status, start time, duration, token count, and tool-call count, with pagination.
  • View on a row expands an inline detail panel with two tabs:
    • Transcript — the full message history as chat bubbles (user, assistant, system, and tool messages with timestamps).
    • Execution trace — the per-step event log of the run: orchestrator decisions, worker spawns, tool calls and results, model calls, and the final response. Each event expands to its raw payload, which answers audit questions like "what exactly did the model see?".

My Sessions (/my-sessions)​

/my-sessions shows every agent run you have started — the "where did my work land" page. Available to every signed-in user.

Runs survive browser disconnects: execution continues server-side even if you close your laptop, files land in your file space, and the finished run appears here with its outputs ready to download.

  • Table view — status (Running, Completed, Error, Parked), agent, intent, a Files column with the artifact count and the name of the top artifact, tool-call count, duration, and a relative start time, with pagination.
  • Auto-refresh — the list polls every 5 seconds while a run is in progress (with an on-screen indicator) and settles to a slower refresh when idle, so completions appear within seconds. A manual Refresh button is also available.
  • Status filter — All, Running, Completed, Error, or Parked.
  • Detail drawer — clicking a row opens a drawer with the run's status, agent, duration, tool and token counts, and:
    • Goal — what the run set out to do.
    • Artifacts — every file or item the run produced, with size and type. Internal files download directly; items created in external systems (for example an issue or a message) open in a new tab.
    • Outcome — completed actions and any pending items.
    • Tool calls — each tool invocation with its status and duration, with logs links into File Space where execution logs were captured.
    • Errors — any errors the run recorded.