---
audience: developer
summary: Specialist AI configurations and the skills library.
title: Helpers and skills API
path: reference/api/helpers-and-skills
status: published
---

# Helpers and skills API

11 endpoints.

## Helpers

A **helper** is a specialist AI exposed as a tool that other AIs
can invoke. Where a sidekick is a delegate, a helper is a one-shot
consult.

| Method | Path | Purpose |
|---|---|---|
| `GET` | `/v1/rooms/{room_id}/helpers` | Helpers available to AIs in this room. |
| `GET` | `/v1/admin/helpers` | All helpers configured for the tenant. |
| `PUT` | `/v1/admin/participants/{participant_id}/helper` | Configure a participant as a helper (or update its helper config). |
| `GET` | `/v1/admin/participants/{participant_id}/helper/invocations` | Invocation log. |
| `POST` | `/v1/admin/participants/{participant_id}/helper/playground` | Test-invoke without storing. |

### PUT .../helper body

```json
{
  "helper_enabled": true,
  "helper_role_description": "You are a senior SRE on-call helper. Diagnose incidents.",
  "helper_temperature": 30,
  "helper_daily_token_cap": 100000,
  "planner_enabled": true,
  "above_threshold_dispatcher_only": false
}
```

`helper_temperature` is `x100` (so `30 = 0.30`) to keep the column an int.

`above_threshold_dispatcher_only=true` means this helper is invoked
via the generic `call_helper(name, query)` tool, not via a dedicated
`ask_<slug>` tool — useful when the tenant has many helpers and you
don't want to bloat every AI's function-calling schema.

## Skills

The skills library is a curated registry of installable / invokable
skills.

| Method | Path | Purpose |
|---|---|---|
| `GET` | `/v1/skills/catalog` | Browse available skills. |
| `POST` | `/v1/skills/{slug}/install` | Install for the tenant. |
| `POST` | `/v1/skills/{slug}/uninstall` | Uninstall. |
| `POST` | `/v1/skills/{slug}/approve` | Approve a "needs approval" skill. |
| `GET` | `/v1/skills/installed` | List installed skills. |
| `GET` | `/v1/skills/{slug}` | Skill detail. |

Skills are like plugins but distributed as packaged units —
`description`, `parameters`, `executor URL` — and registered with
ScaiSkills. The AI sees installed skills in its toolbox as
`skill_<slug>` tools.
