--- title: "STAIR FAQ" group: "Reference" groupOrder: 5 order: 4 --- Short answers to common STAIR questions about agents, tools, models, API access, security and operations, each linking to the full page. ## FAQ ### What is an agent in STAIR? An agent is a configured AI assistant: a model, optional purpose, boundaries and instructions, plus toolsets, skills and sub-agents. You call it by name from STAIR Chat, the API, workflows or schedules. See [AI Agents](/docs/ai-agents/llm.md). ### Which agent fields are required? Only **Agent Name**, **Description**, **Model**, **Owner** and **Manage Access** are required. **Agent Purpose**, **Boundaries** and **Instructions** are optional Markdown fields. See [AI Agents](/docs/ai-agents/llm.md). ### What is the difference between a toolset, a tool and a skill? A toolset connects one MCP server to STAIR, and each tool is one operation that server offers. A skill is a reusable instruction package (a `SKILL.md` plus optional files) that tells an agent how to do a task. See [Toolsets and MCP Tools](/docs/toolsets-and-mcp-tools/llm.md) and [Agent Skills](/docs/agent-skills/llm.md). ### How does an agent delegate work to other agents? Add existing agents on the **Sub Agents** tab and keep the **Delegation** capability on. The agent can then hand tasks to those sub-agents. See [AI Agents](/docs/ai-agents/llm.md). ### How do I add a new MCP server to STAIR? Find the server in the **MCP Catalog** and either **Deploy** it as a server instance and create a toolset on that instance, or use **Add Toolset** for a hosted endpoint. Then add the toolset to an agent. See [MCP Server Catalog](/docs/mcp-server-catalog/llm.md). ### Which LLMs can STAIR use? STAIR connects to model providers of the types Open AI, Open Router, Azure OpenAI, vLLM, Ollama and Custom. An Admin adds the provider on the **Providers** tab of **Models**, then registers its models. See [LLM Models](/docs/llm-models/llm.md). ### How does STAIR estimate the cost of agent runs? Each model can have an **Input price**, **Output price** and **Cached input price** in USD per million tokens. The **Usage** tab of agents and models uses these prices; an empty price means no cost. See [LLM Models](/docs/llm-models/llm.md). ### Can I call a STAIR agent from my own application? Yes. STAIR offers an OpenAI-compatible API: send `POST /ee/v1/chat/completions` with the agent name as `model`, and authenticate with an `aip_sk_` API key. See [OpenAI-Compatible API](/docs/openai-compatible-api/llm.md). ### How do I get an API key? Create a personal key in the user menu under **Connections** → **Tokens**, or ask an Admin for a service account key. Each key is shown once and must have an expiration date. See [API Keys and Service Accounts](/docs/api-keys-and-service-accounts/llm.md). ### Can external MCP clients use STAIR tools and agents? Yes: an MCP Gateway exposes selected toolset tools and agents as one MCP endpoint. MCP clients authenticate with the `STAIR-API-KEY: aip_sk_…` header. See [MCP Gateways](/docs/mcp-gateways/llm.md). ### Who can use a given agent? The agent's ACL decides: users, teams and service accounts with **Execute Access** can use it, and **Read**, **Write** and **Manage Access** control viewing, editing and administration. See [Security and Access Control](/docs/security-and-access-control/llm.md). ### Does the LLM provider see personal data? Not when the agent has **Handles personal data** turned on. STAIR replaces the selected PII entity types with placeholders such as `` before the LLM call, and fails the request if the PII detector is unavailable. See [Security and Access Control](/docs/security-and-access-control/llm.md). ### Where are credentials such as provider API keys stored? In **Secrets**, where values are masked and each secret has its own ACL. Per-user toolset credentials stay with each user under **Connections** → **Toolset**. See [Secrets](/docs/secrets/llm.md). ### How do I find out who changed something? Open **Activity Log**, or choose **Show History** on a record. Each entry shows the person, the operation and the exact changes. See [Security and Access Control](/docs/security-and-access-control/llm.md). ### Can agents run automatically? Yes. A schedule runs an agent or a workflow on a cron expression, in a chosen time zone, between a start and an end date. See [Scheduled Agent Runs](/docs/scheduled-agent-runs/llm.md). ### How do I chain several agents into one process? Build a workflow: a graph of STAIR Agent, Condition, Notification and Action nodes, where later nodes can use earlier nodes' outputs. See [Agent Workflows](/docs/agent-workflows/llm.md). ### How do I test that an agent answers correctly? Create an agent evaluation with scripted requests and evaluators, run it and track the pass rate. **Compare runs** shows what changed between two runs. See [Agent Evaluations](/docs/agent-evaluations/llm.md). ### Can I move agents between STAIR installations? Yes. **Data Transfer** exports selected records, optionally with their dependencies, into a ZIP bundle that another installation imports. See [Data Transfer](/docs/data-transfer/llm.md). ### Where do end users chat with agents? In STAIR Chat. Users pick an agent in the message composer, attach files, organize chats in folders and share a chat by public link. See [STAIR Chat](/docs/stair-chat/llm.md). ## Related - [What is STAIR](/docs/what-is-stair/llm.md) - [STAIR Concepts](/docs/stair-concepts/llm.md) - [AI Agents](/docs/ai-agents/llm.md) - [OpenAI-Compatible API](/docs/openai-compatible-api/llm.md) - [Security and Access Control](/docs/security-and-access-control/llm.md) - [STAIR Console Navigation](/docs/stair-console-navigation/llm.md)