← Admin Panel

AI agents

admin panel · wireframes · ai

Oversight of the whole organization's personal agents: a list with statuses, the one lever being to pause an agent. Configuration is edited only by the owner in Web App.

Platform run limits → Agent Engine
platform_settings · run limits
achilles.local/admin/agents /admin/agents · Owner / Admin
Run step cap
15
How many “think → call a tool → read” steps the agent gets in a single run. A safeguard against looping, not a budget. Default 15; raised to 30 for deep research.
Concurrent runs per platform
4
How many agent runs execute at once across the whole organization. Extra ones wait in the queue rather than being rejected. The real ceiling is set by the provider's TPM tier.
Applies to every agent in the organization
  1. Two platform run limits, shared across the organization: the run step cap and the concurrency of the background lane. The weekly token budget — a spend limit — has moved to the AI usage screen, next to the chat budget and per-person spend.
All agents → Agent Engine
list · all owners
achilles.local/admin/agents /admin/agents · Owner / Admin
All agents
Search by name and owner
Status Owner Schedule 1
AgentOwnerStatusScheduleLast run
1–50 of 86
1 2
  1. 1 Filters — Status · Owner · Schedule facets. Search and filter behavior is the same across all lists.
  2. 2 Clicking the name opens the agent profile (read-only); clicking the owner opens the user card.
  3. 3 ⋯ opens the row menu: open profile · pause / lift pause. Pausing requires confirmation; the admin has no other action over someone else's agent in v1.
  4. 4 Status combines the owner's on/off switch, the administrator lock, the weekly budget, and model availability: active · disabled · paused · admin · weekly limit · model removed · error. The administrator pause is sticky — the owner can't lift it. Agent Engine.
  5. Bulk actions, editing someone else's agent, and an instant abort of a running run — v2.