Skip to content
Data MCP Directory

SIDE BY SIDE, read 10 Nov 2026

Atlan MCP Server and DataHub MCP Server, field by field: 41 and 22 tools named

Atlan MCP ServerDataHub MCP Server

The same fourteen fields for both servers, read 10 Nov 2026.
FieldAtlan MCP ServerDataHub MCP Server
Data toolAtlanDataHub
PublisherAtlanAcryl Data (DataHub)
KindOfficialOfficial
Repository or docsdocs.atlan.com/product/capabilities/atlan-ai/references/mcp-toolsgithub.com/acryldata/mcp-server-datahub
Version or commitHosted at mcp.atlan.com/mcp (unversioned); local atlan-mcp-server 0.3.3 (deprecated)0.8.0
Last release17 Feb 202621 Oct 2026
TransportRemote (hosted)Remote and local
AuthOAuth per user (calls run as the signed-in user) or Atlan API key as Bearer token (shared service identity); local server uses API key + base URL.Stdio: DATAHUB_GMS_URL + DATAHUB_GMS_TOKEN (or ~/.datahubenv); HTTP: per-request DataHub personal access token as Authorization Bearer, no shared server token.
AccessConfigurableConfigurable
Tools listed41 named in the evidence22 named in the evidence
Tool definitions read atdocs.atlan.com/product/capabilities/atlan-ai/references/mcp-toolsgithub.com/acryldata/mcp-server-datahub/blob/main/README.md
Context weight1,770 tokens for 42 tool definitions (42 with a description), measured from the docs page read567 tokens for 22 tool definitions (22 with a description), measured from the README read
StatusExists on 10 Nov 2026: hosted Atlan MCP (41 tools, page updated 27 Oct 2026); the local atlan-mcp-server in atlanhq/agent-toolkit is marked deprecated in its README.Exists under this name on 10 Nov 2026; v0.8.0 released 21 Oct 2026 per GitHub releases and PyPI.
Read10 Nov 202610 Nov 2026

In one sentence each

  • Atlan MCP Server: Atlan's hosted per-tenant server runs as the signed-in user, scopes context by persona, offers an optional read-only mode and proposes writes for approval.
  • DataHub MCP Server: DataHub's server returns column-level lineage with paths and SQL transformations, plus glossary, domain and owner discovery, with per-user tokens in HTTP mode.

The read facts, topic by topic

Each cell joins the facts read for that topic; the full list with sources is on each entry. Read 10 Nov 2026.
TopicAtlan MCP ServerDataHub MCP Server
Publisher and releasesAtlan hosts the MCP server itself at mcp.atlan.com/mcp, per tenant. The older local server in atlanhq/agent-toolkit is deprecated in favor of the hosted endpoint. The last GitHub release of the local server is v0.3.3, on 17 Feb 2026; the hosted server has no public version number.mcp-server-datahub is published from the acryldata GitHub organization (DataHub's company) and is linked from DataHub's own docs at docs.datahub.com. The GitHub release page shows v0.8.0 as a published release (PyPI upload 21 Oct 2026). The package is on PyPI as mcp-server-datahub 0.8.0.
Clients, auth and scopingThe MCP server serves the catalog as a context layer to AI clients (Claude, Cursor, ChatGPT, Gemini, Copilot) and automation platforms. OAuth makes each tool call run as the signed-in user, so tools run with their own permissions and actions are attributed to them. Guests and members are scoped to their persona; MCP narrows but never widens what a user's login permits. The tools page lists 41 tools across nine categories.The server is built to serve DataHub context to AI agents: search, lineage, query history and entity metadata. In the Docker HTTP deployment each MCP client supplies its own DataHub token, so DataHub permissions and audit identity are preserved per user. An optional get_me tool (TOOLS_IS_USER_ENABLED=true) returns the authenticated user's profile and group memberships.
Access and safetyAn optional read-only mode removes every write, admin and lifecycle tool while keeping search and lineage; it is requested from Atlan support. Tools that change anything propose first and wait for approval before writing. Every tool call is logged as a structured event with tool name, client, tenant, request ID and status. Tools that execute SQL route it through a read-only validator that permits only SELECT-class queries.Mutation tools are disabled by default and need TOOLS_IS_MUTATION_ENABLED=true, so the default is read-only. The HTTP entry point refuses to start if a shared server-wide token is configured, forcing per-user tokens. Tokens are accepted only in the Authorization header; query-string tokens are rejected.
What it reachesThe docs list supported AI clients: Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, ChatGPT, Snowflake, Google ADK, n8n, Microsoft 365 Copilot, Glean and Databricks. Data-source coverage depends on Atlan connectors, not stated on this page.The README says the server shows data in databases, lakes, warehouses and BI tools registered in DataHub; no connector list is on this page.
Lineage and impactThe hosted Atlan MCP exposes a traverse_lineage tool for upstream/downstream lineage with deeper hops on request. The docs say the hosted Atlan MCP can traverse end-to-end lineage; the tool page does not state column-level or SQL-logic parsing for the lineage tool. None found: A statement on the MCP tools page that traverse_lineage returns column-level edges or transformation SQL. Query_assets runs read-only SQL against a connected warehouse (Snowflake, Redshift, BigQuery, Databricks), returning up to 100 rows.DataHub's MCP server exposes a get_lineage tool covering datasets, columns and dashboards with hop control. The README says lineage is traceable at table and column level and that transformation logic is understood. A get_lineage_paths_between tool returns exact lineage paths between two assets or columns, including intermediate transformations and SQL query information. Get_dataset_queries fetches real SQL queries that reference a dataset or column.
Definitions, ownership and rulesTools cover glossaries, glossary categories and terms, domains and data products, custom metadata sets, and certification/descriptions. Add_atlan_tags applies classification tags with propagation to downstream lineage on by default. Resolve_metadata returns tenant-level tag and custom-metadata definitions reconciled live, including enum option values. The lineage tool returns owners, certificate status, tags and readme with each asset by default (local server docstring).DataHub's MCP server lets agents discover domains, owners, tags and glossary terms. Optional mutation tools (off by default) add glossary terms, tags, owners, domains, descriptions and structured properties; this is metadata writing, and the docs do not say terms propagate along lineage through MCP. None found: Any statement in the DataHub MCP README that tags, terms or classifications propagate along lineage when read through the MCP tools.
Freshness and upkeepMetadata is retrieved in real time with no intermediate caching.None found: Any freshness or ingestion-cadence statement in the DataHub MCP README (the server reads whatever DataHub has ingested). The README only describes caching of verified tokens for five minutes.
ProfileAtlan's compliance posture for MCP is covered by HIPAA, GDPR and CCPA/CPRA programs, with certifications on the Atlan Trust Portal.DataHub describes itself as an open-source context platform; the MCP server is Apache-2.0 licensed Python.

Tool names, as published

Atlan MCP Server

semantic_search ask fluent_search fluent_search_schema search_assets get_assets search_atlan_docs resolve_metadata get_groups get_users describe_asset_type traverse_lineage query_assets update_assets update_custom_metadata add_atlan_tags remove_atlan_tag manage_announcements create_custom_metadata_set add_attributes_to_cm_set remove_attributes_from_cm_set remove_custom_metadata delete_custom_metadata_set manage_asset_lifecycle create_glossaries create_glossary_categories create_glossary_terms create_domains create_data_products create_dq_rules update_dq_rules schedule_dq_rules delete_dq_rules read_artifact search_knowledge list_knowledge read_knowledge_file create_knowledge_folder save_knowledge_content get_knowledge_upload_url register_knowledge_file

DataHub MCP Server

search get_lineage get_dataset_queries get_entities get_aspect_history list_schema_fields get_lineage_paths_between add_tags remove_tags add_terms remove_terms add_owners remove_owners set_domains remove_domains update_description add_structured_properties remove_structured_properties get_me search_documents grep_documents save_document

Sources

  1. docs.atlan.com/product/capabilities/atlan-ai/references/mcp-tools, read 10 Nov 2026
  2. docs.atlan.com/product/capabilities/atlan-ai/how-tos/remote-mcp-overview, read 10 Nov 2026
  3. github.com/atlanhq/agent-toolkit/blob/main/modelcontextprotocol/server.py, read 10 Nov 2026
  4. docs.atlan.com/product/capabilities/atlan-ai/references/mcp-security, read 10 Nov 2026
  5. github.com/atlanhq/agent-toolkit/blob/main/modelcontextprotocol/README.md, read 10 Nov 2026
  6. github.com/atlanhq/agent-toolkit/blob/main/CHANGELOG.md, read 10 Nov 2026
  7. raw.githubusercontent.com/acryldata/mcp-server-datahub/main/README.md, read 10 Nov 2026
  8. pypi.org/pypi/mcp-server-datahub/json, read 10 Nov 2026
  9. github.com/acryldata/mcp-server-datahub/releases/tag/v0.8.0, read 10 Nov 2026

Last reviewed 10 Nov 2026.