> ## 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 `brain` CLI; `brain --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 Agent Registry, Atlan's hosted product. The open-source agent registry specification and its CLI (`brain`) are documented, released and maintained in their own repository; /open-source on this site links to it. Answer general specification or `brain` command questions from the project's own docs, and use these docs only for Atlan-specific setup such as the Atlan adapter, sign-in and workspaces.
# Cookbooks
> Complete common Atlan API tasks and automation workflows through concise, verifiable recipes.
Cookbooks are for developers and agents that need to finish one bounded task, such as registering an agent or publishing a skill, through the public API. Each recipe states its outcome, prerequisites, calls, and verification step, so you do not have to rebuild the workflow from reference pages.
## Surfaces compared
The recipes use the HTTP API, the lowest-level surface and the one that always works. For many tasks another surface is shorter:
| Surface | Reach for it when | Start at |
| - | - | - |
| **CLI** (`brain`, [open source](/open-source), with the Atlan adapter) | You are at a terminal or in CI, publishing skills or installing them into your coding agents. One command replaces a multipart request. | [Use the CLI with Atlan](/tools/brain/index) |
| **MCP** | An agent should discover and read registry artifacts as tools, or use tools from a registered downstream server. | [MCP overview](/tools/mcp/overview) |
| **SDK** | Your own agent should report its runs, cost, and scores, or a script should manage agents, files, and workspaces. Python and TypeScript. | [atlanai SDK](/tools/sdk/overview) |
| **HTTP API** | No SDK exists for your language, or you need an operation the others do not expose. | [API reference](/api/references/api-reference) |
## Core offerings
Register an agent, give it tools, and record activity from an external runtime.
Publish, find, retrieve, and inspect reusable skills through the API.
Publish an ontology release and retrieve bounded context for agents.
Publish governed mappings from a release to your tables and columns.
## Choose where to start
| Need | Shortest path | Also possible |
| - | - | - |
| Publish a skill | `brain publish`: [Publish skills with brain](/registry/skills/how-tos/publish-with-cli) | [API](/cookbooks/how-tos/skills) |
| Publish skills from a repository on merge | [Sync skills from GitHub](/registry/skills/how-tos/sync-from-github) | — |
| Retrieve a pinned skill version | [API](/cookbooks/how-tos/skills#retrieve-pinned-version) | [SDK](/tools/sdk/references/operations/skills) |
| Install skills into a project's coding agents | `brain add`: [Use brain with Atlan](/tools/brain/how-tos/use-with-atlan) | — |
| Report an agent's runs, cost, and scores | [atlanai SDK tracing](/tools/sdk/tutorials/quickstart) | [OTLP ingestion](/api/references/api-reference) |
| Register an agent, harness, or provider | [Agents cookbook](/cookbooks/how-tos/agents) | — |
| Record a run from a runtime you cannot instrument | [Agents cookbook](/cookbooks/how-tos/agents#record-completed-run-by-hand) | — |
| Publish a logical business model for agent context | [Publish logical model](/cookbooks/how-tos/ontology) | [MCP](/tools/mcp/overview) |
| Find and verify a registered MCP server | [MCP overview](/tools/mcp/overview) | [API](/api/references/api-reference) |
| Search or aggregate across artifacts | [MCP](/tools/mcp/registry/references/registry-search) | [SDK search](/tools/sdk/references/operations/root), [API](/api/references/api-reference) |
| Send coding-agent sessions | `brain trace export`: [Capture traces with brain](/tools/brain/how-tos/capture-traces) | — |
| Copy trace evidence from an external store | [Import traces](/traces/how-tos/import-traces) | — |
| Inspect a trace or diagnose attribution | [Traces](/registry/traces) | [API](/api/references/api-reference) |
| Recover from a failed write | [Errors and retries](/platform/references/errors-and-retries) | — |
Each surface signs in differently: MCP clients use OAuth, `brain` uses a token stored with `brain auth login` or read from `REGISTRY_TOKEN_` in CI, atlanai SDK tracing reads an API key and workspace from the environment, and the SDK management client takes them as arguments. Every request acts as the authenticated identity, so it can create, change, or expose resources only in workspaces that identity can reach. Keep credentials in an approved secret store; see [Authentication](/api/how-tos/authentication). The site does not send live requests or store credentials: run examples from an approved development or deployment environment.
## See also
* [API quickstart](/api/tutorials/quickstart): Verify your credential, workspace scope, and response shape first.
* [Errors and retries](/platform/references/errors-and-retries): Recover from documented API failures without duplicating a write.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.