> ## 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 AI Registry, Atlan's hosted product. The AI registry specification and aireg CLI are maintained separately. /open-source lists publicly reachable signed installers and explains that the source repository and hosted CLI reference currently require GitHub access. Do not claim the repository or its reference docs are publicly accessible. Use these docs for Atlan-specific setup such as the adapter, sign-in and workspaces. # Atlan commands and skills for coding agents > The /atlan-* commands and skills Atlan gives your coding agent: what each does, when to use it, and what it changes. Atlan ships a set of `/atlan-*` commands and skills that teach a coding agent, such as Claude Code, how to publish skills, register agents, upload files, and record outputs in Atlan. They run in your own session, with your identity, and call the Atlan MCP tools on your behalf. ## How your agent gets them The same commands and skills reach your client two ways: * **Atlan MCP endpoint:** When your client is connected to Atlan's MCP endpoint, each command and skill is offered as an MCP prompt named after it, such as `atlan-new-skill`. Their supporting files are offered as MCP resources under `atlan://defaults/`. See [Connect Claude Code](/clients/claude/how-tos/claude-code). * **The `atlanai` plugin:** Atlan's Claude Code plugin marketplace serves an `atlanai` plugin that carries the same files as plugin commands and skills. In Claude Code, MCP prompts appear as slash commands named `/mcp____`. With the server named `atlan`, as in [Connect Claude Code](/clients/claude/how-tos/claude-code), run: ```text theme={null} /mcp__atlan__atlan-new-skill meeting-notes-to-actions "Turn meeting notes into owners and actions. Use after a meeting." ``` Your client lists only the commands for apps your organization has installed. For example, `/atlan-analytics` appears only when Metrics Studio is installed. Every argument is optional. ## Commands | Command | What it does | Use it when | What it changes | | - | - | - | - | | `/atlan-new-skill` `[skill-name] [description]` | Writes a `SKILL.md`, zips the package, uploads it, and creates the skill. | You want the current work saved as a new skill. | Creates a skill and its first version in the workspace you choose. | | `/atlan-update-skill` `[skill-handle-or-name]` | Publishes a new version when the content changed, or updates only the name and description. | A skill you already published needs a change. | Adds a skill version, or edits the skill's name or description. | | `/atlan-link-skill` `[repository path or sync request]` | Sets up ongoing GitHub-to-Atlan sync for a repository's skills. | You want every merge to publish the repository's skills. | Edits files in your repository, including its GitHub Actions workflow. Skills publish later, from CI. | | `/atlan-new-agent` `[agent-name] [what it does]` | Registers an agent record. | You want an agent listed in a workspace. | Creates an agent record. | | `/atlan-link-agent` `[repository path or request]` | Links the skills and agents authored in a repository to Atlan. | You want a repository's agent, skills, and files to resolve to one agent record. | Creates or updates the agent, publishes its selected skills, and writes link configuration into the repository. | | `/atlan-connect-repo` `[repository path or request]` | Registers a repository and links the skills, agents, and scorers that came from it. | You want Atlan to know which repository a record lives in. | Creates a repository record and relationships to existing records. It publishes nothing. | | `/atlan-upload` `[path-to-local-file]` | Uploads a local file and registers it as a file record. | A later step needs a file in Atlan, such as an output. | Creates a file record. | | `/atlan-record-output` `[file-path-or-url] [output name]` | Records a deliverable as an output, either as an uploaded file or as a link. | A run produced a document, page, or file you want in the Library. | Creates an output, and a file record when it uploads one. See [Save outputs to the Library](/registry/agents/how-tos/save-outputs-to-library). | | `/atlan-analytics` `[build \| improve \| eval] [project-slug]` | Builds a talk-to-data (text-to-SQL) agent, or measures and improves one against an eval. | You want an agent that answers business questions from warehouse data. Requires a Snowflake connection in your client. | Creates metrics, saved queries, an agent record, and eval experiments in Atlan, plus a Claude Code subagent file and a `.atlan-analytics/` folder in your directory. | ## Skills The commands point to these skills for the full procedure. Each skill's description says when to use it. | Skill | What it covers | What it changes | | - | - | - | | `atlan-author-skills` | Creating a skill from a package, or publishing a new version of one. | Creates skills and skill versions. | | `atlan-link-skill` | GitHub-to-Atlan skill sync through the repository manifest and GitHub Actions. | Edits repository files. | | `atlan-author-agents` | Registering agents and coding harnesses, one at a time or in bulk, and recording which skills an agent uses. | Creates and updates agent records. | | `atlan-link-agent` | Linking a repository's agents and skills to one agent record. | Creates or updates the agent and writes repository configuration. | | `atlan-record-outputs` | Recording a deliverable as an output, attributing it to the agent and session that made it, and noting the skill it used. | Creates outputs and output versions. | | `atlan-upload-content` | Uploading content too large for a tool call, such as a skill package or a document. | Creates file records. | | `atlan-read-content` | Reading a file or a skill's `SKILL.md` and references held in Atlan. | Nothing. It only reads. | | `atlan-connect-repo` | Registering a repository and linking the records that came from it. | Creates a repository record and relationships. | | `atlan-analytics`, `atlan-analytics-build`, `atlan-analytics-improve`, `atlan-analytics-eval` | Building, evaluating, and improving a talk-to-data agent. | Creates metrics, saved queries, an agent record, and eval experiments. | ## Rules * **Your identity, your permissions.** Each command acts as the signed-in identity and reaches only the workspaces that identity can access. * **Workspace choice.** When you can write to more than one workspace, the command asks which one to use. * **Nothing runs in Atlan.** These commands work in your client's session. Registering an agent records it; it doesn't deploy or run it. * **Published skills still follow release policy.** A skill version created by a command is scanned and released under the workspace's release policy. See [How skill versions and releases work](/registry/skills/concepts/versions). ## Related * [Skill package format](/registry/skills/references/package-format): The rules `SKILL.md` and its package must meet. * [Connect Claude Code](/clients/claude/how-tos/claude-code): Connect the Atlan MCP endpoint. * [Sync skills from GitHub](/registry/skills/how-tos/sync-from-github): What `/atlan-link-skill` sets up. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.