Skip to content
PRESHaiPRESHaiPRESHos Platform

AI config

AI config is the workspace-level model configuration surface: which models agent work runs on, under which provider and gateway, with per-unit price visible. The surface is /ai/config (Models, Providers, and Gateways tabs sharing one surface with a sync-from-gateway action). Per-agent behavior (profile, runtime policy, and where tools, MCP servers, skills, and knowledge attach) lives one level down in Agent config.

  • Each row carries name, key, provider, capability badges, context window, per-unit price, and status. Inactive rows stay visible so deprecations are deliberate.
  • Capability badges name what the model does: Reasoning, Tools, Vision, Files, Web search, Text.
  • Search by name, key, or URL; sort, columns, and filters narrow the list. Provider and Capability chips filter the same list the tabs slice.
AI config: model rows with provider, capability badges, context window, price, and status
  • Model choice does not change the three guarantees (isolation, approvals, audit). PRESHos sits between agents and the stack; the model is the engine, not the governance.
  • Building or changing an agent starts with Agent Builder, not by editing a running agent.

Treat /ai/config as the source of truth for workspace model config. Selection guidance: Model registry. Routes models are reached through: Gateways. Credentials behind them: AI providers.