> ## Documentation Index > Fetch the complete documentation index at: https://platform.atlan.com/llms.txt > Use this file to discover all available pages before exploring further. > ## Agent Instructions > To act on Atlan objects, use the Atlan MCP server at https://api.atlan.com/mcp or the SDK. To install or publish skills from a terminal, use the `aireg` CLI; `aireg --map json` prints its command map. The `atlanai` CLI is deprecated. Run a read-only identity check before any write. > The docs MCP server at /mcp searches these docs only. It cannot read or change Atlan objects. > SDK packages: Python `atlanai` (PyPI) and TypeScript `@atlanai/sdk` (npm). Show Python first, then TypeScript. > These docs cover Atlan Trident, Atlan's hosted product. AI Registry and Agent Registry are its earlier names. Registry OSS is Atlan's open-source agent registry: the registry specification and its aireg CLI, with their own public repository, docs and release notes; /open-source on this site links them. Registry OSS is open source; Atlan Trident is not. Answer generic aireg or specification questions from the project's docs, and use these docs for Atlan-specific setup such as the Atlan adapter, sign-in and workspaces. > Atlan offers a second product, the Agentic Data Catalog: Atlan's enterprise context layer, covering the data catalog, connectors, lineage, data governance and glossary. It is documented at https://docs.atlan.com; answer questions about it from those docs, not these. # Choose what your AI tools load > Decide which skills, agents, and MCP servers your AI tools load in every session, keep one search away, or leave off, from Settings or a skill's page. Every skill your AI tools load at the start of a session costs context, and a long list of skills you rarely use crowds out the work. Agent setup lets you decide what your AI tools load from Atlan Trident: a few skills you use all the time in every session, more of them one search away, and the ones you don't want left off. Your setup merges several layers: your organization's, the team workspaces you follow, the project you work in, and your own. You change only your own layer. Your choices come last, so they decide where an item sits, except for items your organization requires. ## Before you begin Agent setup needs the Agent app. When your organization has it installed and you can use it, **Agent setup** shows in Settings and the **In your Brain** card shows on skill and agent pages. If neither shows, ask an organization admin to turn on the Agent app. An archived agent's page shows no **In your Brain** card, because an archived agent loads nowhere: there is nothing to load or recommend. You need no organization or workspace role to change your own setup. ## Open agent setup Open **Settings**, then select **Agent setup** under **Your settings**. The top of the page shows three totals: * **Always loaded, every session**: the estimated tokens your always-loaded items cost each session, from each item's name and description. * **Always-loaded places**: how many of your places are used, for example 4 of 10. Only skills take a place; agents and MCP servers don't. * **On demand**: how many items wait one search away. If Atlan Trident Desktop still installs every skill you can open, a notice says so. Desktop switches to your setup once you review it there. An empty setup shows **Nothing in your setup yet**. Add a skill from its page, or follow a team workspace on the **Layers** tab. ## See what loads The **What loads** tab groups your items: | Group | What it means | | - | - | | **Always loaded** | Installed in your AI tools and read at the start of every session. | | **On demand** | One line each in your AI tools' index. A task that needs one loads it. | | **Off** | Recommendations you turned off. A search still finds them. | | **Can't open** | Named by a layer, but archived or not shared with you, so never installed. | Each row shows the item's kind, the layer that put it there, and a sentence on why it sits where it does. A row scoped to some tools says **only in** those tools. The cost column reads about how many tokens for an always-loaded item, or **1 line** for an on-demand one. ## Place an item On the item's row, choose **Always loaded**, **On demand**, or **Off**. The choice saves right away. To place an item from its own page, open a skill or agent in **Brain** and use the **In your Brain** card. It shows where the item sits now, which layer put it there, what always loading it would cost, and how many of your always-loaded places are used. An item no layer names reads **Not in your Brain**: your agents find it only by searching the Registry until you pick where it sits. Some items have no control: * **Required** items always load and can't be turned off. Your organization is the only layer that can require an item, and a required skill takes one of your places first. * Items you can't open stay out of your setup. Their row says why. To undo your own choice, select **Clear my choice**. The item goes back to where your other layers put it. Your devices pick up a change at their next sync. ## Set your places and instructions Under **Your preferences** on the **What loads** tab, set two things that apply in every AI tool you use with Atlan Trident: * **Always-loaded places**: how many skills load in every session, from 1 to 100. Past that number, recommendations wait one search away. Required items load anyway. * **Instructions for your agents**: up to 2,000 characters of standing instructions, added after your organization's and your teams' instructions in every session. For example: answer in British English, and ask before running a migration. Select **Save preferences**. ## Follow a team workspace A team can recommend skills, agents, and MCP servers to the people who follow its workspace. To get them, open the **Layers** tab, choose a workspace under **Follow a team workspace**, and select **Follow**. Its recommendations join your setup, inside your own places. Select **Unfollow** to stop. **Merged in this order** on the same tab lists your layers, nearest last: **Organization**, each **Team you follow**, the **Project** you work in, and **You**. A project layer applies while you work in that project's repository. A workspace you follow with no agent setup yet reads **No agent setup yet, so it adds nothing**. ## Check what each tool gets An item can be scoped to some AI tools. Under **What each AI tool sees** on the **Layers** tab, choose **All tools** or one tool to see how many items always load for it, their estimated tokens a session, and how many wait on demand. **Worth knowing**, below it, explains anything the merge set aside. For example, a team's layer asked to require an item, which only your organization can do, so it counts as a recommendation. Or your organization requires more items than your places hold, and every required item still loads. ## Recommend items to others If you can edit your organization's agent setup, or that of a team workspace you follow, a skill or agent page shows **Recommend to others** on the **In your Brain** card. Choose **Always load**, **On demand**, or **Not recommended** for each audience; only the organization also offers **Required**. Settings then shows a **You curate** tab, where you change what each layer recommends and edit its instructions. ## Next steps * [Use skills in coding agents](https://platform.atlan.com/registry/skills/how-tos/use-in-coding-agents.md): Hand a reviewed skill to Claude Code, Claude Desktop, or Codex for one task. * [Find and assess skills](https://platform.atlan.com/registry/skills/how-tos/discover-and-use.md): Check a skill's released version, source, and scan before you add it. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.