Atlan

- Search for assets— Find data assets using conditions withsearch_assets.
- Retrieve assets via DSL— Fetch assets using a domain-specific language query withget_assets_by_dsl.
- Explore data lineage— Trace upstream and downstream relationships for an asset usingtraverse_lineage.
- Update asset metadata— Modify user descriptions and certificate status on assets viaupdate_assets.
- Manage business glossaries— Create glossaries, categories, and terms usingcreate_glossaries,create_glossary_categories, andcreate_glossary_terms.
- Govern data quality rules— Create, update, schedule, and delete data quality rules withcreate_dq_rules,update_dq_rules,schedule_dq_rules, anddelete_dq_rules.

[!WARNING]This local MCP server is deprecated.Use the hosted Atlan MCP atmcp.atlan.com/mcpinstead.

The local install path (Docker, uv, orpip install atlan-mcp-server) is in maintenance-only mode — no new features, support not guaranteed. The hosted endpoint is the recommended way to integrate Atlan with Claude Desktop, Cursor, Codex, Databricks UC, and other MCP clients. See theAtlan MCP overviewfor setup.

The AtlanModel Context Protocolserver allows your AI agents to interact with Atlan services.
- Generate Atlan API key by following thedocumentation.
- Select one of the following approaches based on your preference:

- Install via Docker- Uses Docker containers (recommended)
- Install via uv- Uses UV package manager

[!NOTE] Make sure to replace<YOUR_API_KEY>,<YOUR_INSTANCE>, and<YOUR_AGENT_ID>with your actual Atlan API key, instance URL, and agent ID(optional) in the configuration file respectively.

- Follow the officialDocker installation guidefor your operating system
- Verify Docker is running:

docker --version

{ "mcpServers": { "atlan": { "command": "docker", "args": [ "run", "-i", "--rm", "-e", "ATLAN_API_KEY=<YOUR_API_KEY>", "-e", "ATLAN_BASE_URL=https://<YOUR_INSTANCE>.atlan.com", "-e", "ATLAN_AGENT_ID=<YOUR_AGENT_ID>", "ghcr.io/atlanhq/atlan-mcp-server:latest" ] } } }

OpenCursor > Settings > Tools & Integrations > New MCP Serverto include the following:

{ "mcpServers": { "atlan": { "command": "docker", "args": [ "run", "-i", "--rm", "-e", "ATLAN_API_KEY=<YOUR_API_KEY>", "-e", "ATLAN_BASE_URL=https://<YOUR_INSTANCE>.atlan.com", "-e", "ATLAN_AGENT_ID=<YOUR_AGENT_ID>", "ghcr.io/atlanhq/atlan-mcp-server:latest" ] } } }

- Install uv:

# macOS/Linux curl -LsSf https://astral.sh/uv/install.sh | sh # Windows (PowerShell) powershell -c "irm https://astral.sh/uv/install.ps1 | iex" # Alternative: if you already have Python/pip pip install uv

[!NOTE] With uv,uvxautomatically fetches the latest version each time you run it. For more predictable behavior, consider using the Docker option.

Go toClaude > Settings > Developer > Edit Config > claude_desktop_config.jsonto include the following:

{ "mcpServers": { "atlan": { "command": "uvx", "args": ["atlan-mcp-server"], "env": { "ATLAN_API_KEY": "<YOUR_API_KEY>", "ATLAN_BASE_URL": "https://<YOUR_INSTANCE>.atlan.com", "ATLAN_AGENT_ID": "<YOUR_AGENT_ID>" } } } }

OpenCursor > Settings > Tools & Integrations > New MCP Serverto include the following:

{ "mcpServers": { "atlan": { "command": "uvx", "args": ["atlan-mcp-server"], "env": { "ATLAN_API_KEY": "<YOUR_API_KEY>", "ATLAN_BASE_URL": "https://<YOUR_INSTANCE>.atlan.com", "ATLAN_AGENT_ID": "<YOUR_AGENT_ID>" } } } }

The Atlan MCP Server includes a configurable tool restriction middleware that allows you to control which tools are available to users. This is useful for implementing role-based access control or restricting certain operations in specific environments.

You can restrict access to specific tools using theRESTRICTED_TOOLSenvironment variable. Provide a comma-separated list of tool names that should be blocked:

{ "mcpServers": { "atlan": { "command": "docker", "args": [ "run", "-i", "--rm", "-e", "ATLAN_API_KEY=<YOUR_API_KEY>", "-e", "ATLAN_BASE_URL=https://<YOUR_INSTANCE>.atlan.com", "-e", "ATLAN_AGENT_ID=<YOUR_AGENT_ID>", "-e", "RESTRICTED_TOOLS=get_assets_by_dsl_tool,update_assets_tool", "ghcr.io/atlanhq/atlan-mcp-server:latest" ] } } }

{ "mcpServers": { "atlan": { "command": "uvx", "args": ["atlan-mcp-server"], "env": { "ATLAN_API_KEY": "<YOUR_API_KEY>", "ATLAN_BASE_URL": "https://<YOUR_INSTANCE>.atlan.com", "ATLAN_AGENT_ID": "<YOUR_AGENT_ID>", "RESTRICTED_TOOLS": "get_assets_by_dsl_tool,update_assets_tool" } } } }

You can restrict any of the following tools:

- search_assets_tool- Asset search functionality
- get_assets_by_dsl_tool- DSL query execution
- traverse_lineage_tool- Lineage traversal
- update_assets_tool- Asset updates (descriptions, certificates)
- create_glossaries- Glossary creation
- create_glossary_categories- Category creation
- create_glossary_terms- Term creation
- create_dq_rules_tool- Data quality rule creation
- update_dq_rules_tool- Data quality rule updates
- schedule_dq_rules_tool- Data quality rule scheduling
- delete_dq_rules_tool- Data quality rule deletion

RESTRICTED_TOOLS=update_assets_tool,create_glossaries,create_glossary_categories,create_glossary_terms,create_dq_rules_tool,update_dq_rules_tool,schedule_dq_rules_tool,delete_dq_rules_tool

RESTRICTED_TOOLS=get_assets_by_dsl_tool,update_assets_tool,traverse_lineage_tool,create_glossaries,create_glossary_categories,create_glossary_terms,create_dq_rules_tool,update_dq_rules_tool,schedule_dq_rules_tool,delete_dq_rules_tool

- Hidden from listings: Restricted tools won't appear when clients request available tools
- Execution blocked: If someone tries to execute a restricted tool, they'll receive a clear error message
- Logged: All access decisions are logged for monitoring and debugging

If you don't set theRESTRICTED_TOOLSenvironment variable, all tools will be available by default.

The Atlan MCP Server supports three transport modes, each optimized for different deployment scenarios. For more details about MCP transport modes, see theofficial MCP documentation.

For comprehensive deployment instructions, configuration examples, and production best practices, see ourDeployment Guide.

- Host the Atlan MCP container image on the cloud/platform of your choice
- Make sure you add all the required environment variables
- Choose the appropriate transport mode for your deployment scenario. SSE Transport is recommended for production (-e MCP_TRANSPORT=sse)
- For detailed deployment scenarios and configurations, refer to theDeployment Guide

You can use themcp-remotelocal proxy tool to connect it to your remote MCP server.

{ "mcpServers": { "math": { "command": "npx", "args": ["mcp-remote", "https://hosted-domain"] } } }

Want to develop locally? Check out ourLocal BuildGuide for a step-by-step walkthrough!

- Reach out tosupport@atlan.comfor any questions or feedback
- You can also directly create aGitHub issueand we will answer it for you

Technical details: The Atlan MCP server is implemented as a Python application. The Model Context Protocol itself is language-agnostic, but our current implementation requires Python 3.11+ to run.
- If Claude Desktop shows an error similar tospawn uv ENOENT {"context":"connection","stack":"Error: spawn uv ENOENT\n at ChildProcess._handle.onexit, it is most likelythisissue where Claude is unable to find uv. To fix it:

Official MCP server for dbt (data build tool) providing integration with dbt Core/Cloud CLI, project metadata discovery, model information, and semantic layer querying capabilities.