> For the complete documentation index, see [llms.txt](https://docs.euno.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.euno.ai/ai/mcp-setup.md).

# MCP Setup Guide

The Euno AI Assistant can connect to compatible AI clients through the Model Context Protocol (MCP). This lets you explore your data model from Cursor, Claude, ChatGPT, and other MCP-enabled assistants.

The standard connection is `https://api.app.euno.ai/mcp` and uses OAuth. The first time you use Euno MCP, your client prompts you to sign in to Euno in a browser. If Euno cannot identify one clear persona, you will be asked to choose one.

This guide covers the standard direct MCP setup using OAuth.

## Prerequisites

Before setting up MCP, make sure you have:

* access to Euno and at least one assigned persona
* an MCP-compatible client such as Cursor, Claude, ChatGPT in Developer Mode, or Claude Code in VS Code

## Sign In with OAuth

In the recommended OAuth setup, your client will prompt you to sign in to Euno in a browser the first time you connect.

## Personas in MCP

Euno MCP uses your active persona to determine what data and actions are available.

* If Euno can identify which persona to use, it will select it automatically.
* This usually happens when you have a default persona, or only one available persona in the relevant account.
* If Euno cannot identify a single clear persona, you will be asked to choose one.

When `switch_persona` succeeds on an unscoped OAuth connection, Euno stores one active persona for your signed-in user. The saved choice is shared across that user's unscoped OAuth clients, sessions, and threads that can access the selected account. Explicit account and persona settings in a connection override this stored choice and cannot be changed with `switch_persona`. API-key users must change their configured account and persona scope instead.

Switching persona changes the account and permissions used by later unscoped MCP calls.

## Choose Your Setup

Use the setup mode that matches your client's MCP capabilities:

| Use this when                                                                                | Recommended setup                                                                       |
| -------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------- |
| Your client supports remote HTTP MCP servers                                                 | [Direct connection](https://docs.euno.ai/ai/mcp-setup/direct-connection.md)             |
| You use **Google Gemini Enterprise** (Custom MCP Server data store)                          | [Gemini Enterprise Setup](https://docs.euno.ai/ai/mcp-setup/gemini-enterprise-setup.md) |
| Your client does not support remote HTTP MCP servers, or you prefer a local stdio MCP server | [MCP Bridge Setup](https://docs.euno.ai/ai/mcp-setup/mcp-bridge-setup.md)               |
| A non-Claude client cannot complete interactive OAuth sign-in                                | [Advanced API Key Setup](https://docs.euno.ai/ai/mcp-api-key-setup.md)                  |

Claude and Anthropic Directory connections use the direct OAuth setup.

## Read and Write Behavior

Euno MCP includes both read and write tools. Most tools only retrieve Euno data. The following tools change account state, so your client may ask for confirmation:

* `switch_persona` saves the OAuth user's shared active persona. It works only for unscoped OAuth connections.
* The catalog create, update, and delete tools change glossary terms, metadata tags, and workflows.
* The two audit tools may initialize or update Euno's audit storage before returning read results.
* The context benchmark write tools create, update, or delete benchmark questions, start or stop benchmark runs, change the benchmark schedule, and prepare, approve, undo, or reapply suggested context changes. Approving or reapplying a suggestion saves customer skills, glossary entries, or metadata tags.

Euno still checks authentication and persona permissions on the server. Client confirmation does not replace these checks. The MCP server keeps no session state, so a connection keeps working across server restarts and scaling.

## Available Tools

| Tool                                               | Description                                                                                         | Availability                                      |
| -------------------------------------------------- | --------------------------------------------------------------------------------------------------- | ------------------------------------------------- |
| `load_account_settings`                            | Return current account, persona, skill, and metadata context.                                       | Always                                            |
| `list_personas`                                    | List accounts and personas with numbered IDs based on the current result.                           | Always                                            |
| `current_persona`                                  | Return the effective account and persona, including a shared OAuth selection.                       | Always                                            |
| `switch_persona`                                   | Save the OAuth user's shared active choice.                                                         | Always visible; succeeds only with unscoped OAuth |
| `load_skill`                                       | Return an Euno skill's instructions and recommended workflow.                                       | Always                                            |
| `search_properties`                                | Search property definitions available to the active persona.                                        | Always                                            |
| `fetch_single_resource`                            | Return one resource by URI with requested properties and relationships.                             | Always                                            |
| `find_resource_by_name`                            | Find resources by exact, partial, or similar name.                                                  | Always                                            |
| `find_resources_for_topic`                         | Find resources related to supplied topics, descriptions, synonyms, or tags.                         | Always                                            |
| `get_upstream_lineage`                             | Return upstream dependencies for one resource URI.                                                  | Always                                            |
| `resource_impact_analysis`                         | Return downstream impact for one resource URI.                                                      | Always                                            |
| `query_to_eql`                                     | Convert a natural-language resource filter into EQL without executing it.                           | Always                                            |
| `search_resources_eql`                             | Search resources with EQL, pagination, sorting, properties, and relationships.                      | Always                                            |
| `count_resources_eql`                              | Count resources with EQL.                                                                           | Always                                            |
| `multi_count_resources_eql`                        | Count resources for up to 200 EQL filters in one call, returned in input order.                     | Always                                            |
| `pivot_resources_eql`                              | Group resources matching EQL by one or more properties and count or aggregate a property per group. | Always                                            |
| `search_resource_audit_log`                        | Return audit events; requires `uri`, `actor`, or `property_id`.                                     | Always                                            |
| `sensitive_data_incident_response`                 | Return evidence for one user/BI-resource pair or a batch of pairs.                                  | Always                                            |
| `documentation_search`                             | Search indexed Euno User Manual pages.                                                              | Always                                            |
| `documentation_get_full_document`                  | Return the text of an indexed Euno User Manual page.                                                | Always                                            |
| `documentation_get_surrounding_context`            | Return adjacent manual chunks for a search result.                                                  | Always                                            |
| `list_metadata_tags`                               | List metadata tags visible to the active persona.                                                   | Always                                            |
| `list_metadata_tag_domains`                        | List the domains where the active persona can create metadata tags.                                 | Always                                            |
| `preview_metadata_tag_plan`                        | Validate a tag create or update without saving it, with scope count, samples, and option counts.    | Always                                            |
| `get_context_benchmark_overview`                   | Return per-domain benchmark scores, active-run state, and the schedule.                             | Account admin persona                             |
| `get_context_benchmark_questions`                  | Return a domain's benchmark questions and suite revision.                                           | Account admin persona                             |
| `list_context_benchmark_runs`                      | List benchmark runs without per-question detail.                                                    | Account admin persona                             |
| `get_context_benchmark_run`                        | Return one benchmark run with per-question summaries.                                               | Account admin persona                             |
| `get_context_benchmark_improvement_report`         | Return the report and rounds of a run that applied suggestions automatically.                       | Account admin persona                             |
| `get_context_benchmark_question_result`            | Return a question's trials, grading, root cause, and suggestion.                                    | Account admin persona                             |
| `get_context_benchmark_suggestion_application`     | Return a suggestion's application state and proposed changes.                                       | Account admin persona                             |
| `get_context_benchmark_schedule`                   | Return the benchmark schedule.                                                                      | Account admin persona                             |
| `create_context_benchmark_question`                | Add a benchmark question to a domain.                                                               | Account admin persona                             |
| `update_context_benchmark_question`                | Change a benchmark question.                                                                        | Account admin persona                             |
| `delete_context_benchmark_question`                | Delete a benchmark question; requires `confirm=true`.                                               | Account admin persona                             |
| `start_context_benchmark_run`                      | Start a benchmark run for some or all domains.                                                      | Account admin persona                             |
| `stop_context_benchmark_run`                       | Stop a benchmark run.                                                                               | Account admin persona                             |
| `update_context_benchmark_schedule`                | Enable, change, or disable the benchmark schedule.                                                  | Account admin persona                             |
| `prepare_context_benchmark_suggestion_application` | Prepare a suggestion as concrete context changes for review.                                        | Account admin persona                             |
| `decide_context_benchmark_suggestion_application`  | Approve or reject a prepared suggestion.                                                            | Account admin persona                             |
| `undo_context_benchmark_suggestion`                | Revert an applied suggestion.                                                                       | Account admin persona                             |
| `reapply_context_benchmark_suggestion`             | Save again a suggestion that an automatic run tested and restored.                                  | Account admin persona                             |

Every tool is listed from the start of a connection. The `tags` and `context_benchmark` skills describe the recommended workflow for their tools; load them with `load_skill` before working with metadata tags or the context benchmark.

Euno accepts the context benchmark tools only for an account admin persona. Benchmark runs and suggestion preparation continue in the background, so their tools return immediately and the read tools report progress.

## Example Prompts

* "Find the Tableau dashboard named Revenue Overview and show its upstream lineage."
* "What dashboards and owners would be affected if resource URI `<resource-uri>` changed?"
* "Count Snowflake tables without an owner."
* "Search the Euno documentation for persona switching and return the full matching page."
* "List our metadata tags and update the certification tag's description."
* "Load the context benchmark skill, show the Finance domain's lowest-scoring benchmark questions, and explain why they failed."

## Help and Privacy

* [Live Euno MCP guide](https://docs.euno.ai/ai/mcp-setup)
* [Direct-connection troubleshooting](https://docs.euno.ai/ai/mcp-setup/direct-connection#troubleshooting)
* [Euno website privacy policy](https://euno.ai/privacy-policy)
* [Euno support resources](https://docs.euno.ai/support-resources) or <support@euno.ai>

### Catalog Tools

Euno MCP also provides 14 direct catalog tools:

* Glossary: `fetch_glossary_term`, `find_glossary_term`, `create_glossary_term`, `update_glossary_term`, and `delete_glossary_term`
* Metadata tags: `fetch_metadata_tag`, `create_metadata_tag`, `update_metadata_tag`, and `delete_metadata_tag`
* Workflows: `fetch_workflow`, `find_workflow`, `create_workflow`, `update_workflow`, and `delete_workflow`

MCP publishes catalog reads as read-only. Catalog creates, updates, and deletes are marked destructive so Claude asks for confirmation before an operation that modifies data. Euno still checks your active persona and backend permissions for every call.

All delete tools require `confirm=true`. Without it, Euno returns a confirmation-required result and does not make the delete request.

`preview_metadata_tag_plan` validates a tag create or update without saving it and shows how many assets the tag covers, with samples and option counts. `create_metadata_tag` and `update_metadata_tag` save the tag immediately. Euno validates the tag first: an invalid tag returns `valid=false` with its errors and nothing is saved. The `tags` skill describes the recommended workflow, including how to copy a fetched tag's full definition into an update.

See [Supported Tags and Previews](https://docs.euno.ai/using-euno/metadata-activation/metadata-tags/creating-metadata-tags-with-ai.md#supported-tags-and-previews) for supported tag types.

`count_resources_eql` can check how many resources match a workflow EQL filter. It does not simulate the complete workflow. Catalog writes are not exposed through the legacy MCP endpoints or the Snowflake batch adapter.

Successful catalog calls return the backend route's JSON shape unchanged. Failures are MCP tool errors with the shared `status`, `error_type`, `status_code`, `body_format`, and `body` fields. If `status` is `outcome_unknown`, check the current catalog state before deciding whether to retry.

***


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.euno.ai/ai/mcp-setup.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
