# Bonnard vs Tableau: MCP server features compared

> Bonnard and Tableau MCP servers compared: what the agent knows about your data, how wrong answers get fixed, who can connect and pricing.

With Tableau (tableau.com), the agent queries a published data source through its fields and calculations, with no SQL. With Bonnard (bonnard.ai), the agent writes SQL on your database from your schema and your team's context.

## Five differences

### What the agent knows about your data

Bonnard gives its context to the AI client the person already uses. `explore_schema` searches table names, descriptions, column comments, enum labels, metrics and business terms, and returns up to 10 tables. For each table the agent gets the keys, row count and sample rows from the database, plus what your team adds: descriptions, joins, metrics, business terms and the values of columns you mark as searchable ([context](/features/context-engine)). A metric is its calculation in SQL, with instructions. The agent reads it and writes the query. Bonnard does not rewrite the query to apply it.

Tableau also gives its context to the AI client. `get-datasource-metadata` returns a published data source's description, its logical tables and their relationships, each field's role, type and default aggregation, the formula of each calculated field, and parameters ([get-datasource-metadata](https://tableau.github.io/tableau-mcp/docs/tools/data-qna/get-datasource-metadata)). `search-content` searches 11 content types and shows whether a data source is certified ([search-content](https://tableau.github.io/tableau-mcp/docs/tools/content-exploration/search-content)). The Pulse tools list Pulse metric definitions ([tools](https://tableau.github.io/tableau-mcp/docs/intro)).

On the Cloud+ edition and the Tableau+ Bundle, Tableau Knowledge adds a knowledge graph of the site's content ([Tableau pricing](https://www.tableau.com/pricing/cloud)). A Creator adds business preferences to it: instructions in plain language about terms, rules and how to use the data ([Tableau Knowledge](https://help.tableau.com/current/online/en-us/tableau_knowledge_about.htm)). `query-knowledge-context` reads those definitions, plus relationships and lineage ([query-knowledge-context](https://tableau.github.io/tableau-mcp/docs/tools/knowledge/query-knowledge-context)).

Tableau's agent does not write SQL. `query-datasource` takes a VizQL query, a list of fields and filters, and runs it on a published data source ([query-datasource](https://tableau.github.io/tableau-mcp/docs/tools/data-qna/query-datasource)). The joins and calculated fields of the data source apply to every query. The agent can still add its own calculation to a query, so a metric defined in Tableau is used when the agent selects that field.

### How wrong answers get found and fixed

Bonnard runs a loop of four steps on how agents use your data. Tableau logs agent use and, on Cloud+, suggests fixes from the state of its knowledge graph, but it has no draft step and no evals for your own questions.

[Image: The context loop in four steps: observe, diagnose, improve, verify.]

**Observe.** Bonnard records each tool call with its SQL, status and error, from every AI client your people use.

Tableau records agent use. Hosted Tableau MCP logs sign-ins and tool calls through Tableau's activity and audit logs ([hosted Tableau MCP](https://tableau.github.io/tableau-mcp/docs/hosted-tableau-mcp)). A self-hosted server writes its own logs and can send metrics to a telemetry provider you supply ([Tableau Server guide](https://tableau.github.io/tableau-mcp/docs/enterprise/tableau-server)).

**Diagnose.** Bonnard reads that record and finds what is wrong: definitions that differ between people, searches that found nothing, sessions that took many calls, repeated errors and tables that stopped updating. Where the data cannot settle a question, Bonnard asks your data team.

[Image: Bonnard's activity record over 30 days. It finds three definitions of net revenue across eight people, a search for gift cards that found nothing, sessions that took nine calls, an undefined column error and a table that stopped in May.]

On Cloud+ and the Tableau+ Bundle, `inspect-knowledge-context` returns a health score for the knowledge graph and a list of suggestions ([inspect-knowledge-context](https://tableau.github.io/tableau-mcp/docs/tools/knowledge/inspect-knowledge-context)). The suggestions name assets with no description, owner or certification, assets with no connections and business context that is incomplete ([Tableau Knowledge](https://help.tableau.com/current/online/en-us/tableau_knowledge_about.htm)). They describe the state of the graph. Ask Tableau whether any of them use agent conversations.

**Improve.** Your data team's answer goes into a draft of the context.

`manage-knowledge-context` creates, changes and deletes business preferences in the shared graph, with no draft step ([manage-knowledge-context](https://tableau.github.io/tableau-mcp/docs/tools/knowledge/manage-knowledge-context)). Workbooks and data sources keep a revision history, with a revision saved at each publish ([revision history](https://help.tableau.com/current/pro/desktop/en-us/qs_revision_history.htm)).

**Verify.** Evals run the same questions against the draft and the live version, and show which answers got better. A person then publishes the draft, and Bonnard keeps each version.

[Image: Three questions, each run five times against the live context and against the draft. Two get better and one shows no clear difference.]

Tableau MCP has no eval feature for your own questions. The eval tests in its repository grade the tools for Tableau's developers, on Tableau's test site ([eval tests](https://tableau.github.io/tableau-mcp/docs/developers/eval-tests)).

`query-datasource` checks a query as the agent works. It sends extra requests that examine string filter values, so the agent can correct a malformed query. A TOP filter combined with dimension filters that are not context filters gets a warning, and a result that a row limit cuts says so.

Two warnings also reach the agent as it works. A result over the row limit tells the agent not to sum or average it. On PostgreSQL, a join that could count rows twice gets a warning.

Tableau's suggestions come from what the graph lacks. Bonnard's come from what agents asked and where they failed, and evals test each change before it is published.

### What you need first

Bonnard starts from your database. Add it as a data source through a read-only credential, then give people access. The agent reads the table comments, keys and enum values that the database declares, and the context your team adds in Bonnard.

Tableau starts from data sources published to Tableau Cloud or Tableau Server. VizQL Data Service must be on, and each person needs the API Access capability on each data source that the agent queries ([configuring Tableau](https://tableau.github.io/tableau-mcp/docs/configuration/tableau-config)). VizQL Data Service does not accept cube data sources ([VizQL Data Service limits](https://help.tableau.com/current/api/vizql-data-service/en-us/docs/vds_limitations.html)). A data source embedded in a workbook can be queried on Tableau 2026.3 or later, once a site admin allows it ([query-datasource](https://tableau.github.io/tableau-mcp/docs/tools/data-qna/query-datasource)). The context is as good as the descriptions, relationships and calculated fields your team has written in those data sources.

### Who can connect

A Bonnard workspace has unlimited users, your customers' users included. Each customer's query runs as that customer's tenant (the value that marks its data in your database). A customer with no tenant gets no data. Isolation uses the database: row-level security on PostgreSQL, Redshift and SQL Server, or one dataset or database for each tenant on BigQuery and Azure SQL Database ([tenancy](/docs/access/tenancy)).

On Tableau, each person signs in with their own Tableau Cloud account through OAuth 2.1. The server then calls Tableau as that person, so their permissions apply ([hosted Tableau MCP](https://tableau.github.io/tableau-mcp/docs/hosted-tableau-mcp)). Creator, Explorer and Viewer licences all include hosted Tableau MCP ([Tableau pricing](https://www.tableau.com/pricing/cloud)). A customer's user connects only as a licensed user of your Tableau site.

### How you pay

Bonnard bills for each call, with no seats or minimums. Free includes 3,000 calls a month, then USD 0.95 for each 100 calls once an admin turns on extra usage ([pricing](/pricing)).

Tableau Cloud bills for each user, on an annual contract, and each deployment needs at least one Creator licence. All editions include Tableau MCP. Standard is USD 75 for each Creator a month, USD 42 for each Explorer and USD 15 for each Viewer. Enterprise is USD 115, USD 70 and USD 35 for the same licences. Tableau Knowledge needs Cloud+ or the Tableau+ Bundle, and Tableau's sales team quotes both. Capacity-based pricing gives Viewer access with no licence for each viewer, also quoted by sales ([Tableau pricing](https://www.tableau.com/pricing/cloud)).

Queries from `query-datasource` go to VizQL Data Service. Each Creator licence raises the site's cap by 100 queries an hour ([VizQL Data Service limits](https://help.tableau.com/current/api/vizql-data-service/en-us/docs/vds_limitations.html)).

## Side by side

Table: Bonnard and Tableau MCP servers, verified 16 Nov 2026

| Feature | Bonnard | Tableau |
| --- | --- | --- |
| Who answers | The AI client the person uses, with Bonnard's context | The AI client the person uses, with Tableau's metadata |
| Context the agent reads | Schema, descriptions, joins, metrics, business terms, searchable values | Data source descriptions, relationships, fields, calculations, parameters, Pulse metric definitions. Business preferences on Cloud+ |
| Context through MCP | `explore_schema`, for any AI client | `get-datasource-metadata`, for any AI client. `query-knowledge-context` on Cloud+ |
| Metric definitions | The agent reads the SQL and writes the query | Queries run through the data source's fields and calculations, with no SQL. The agent can add its own calculation |
| Observe | Records each tool call, from any AI client | Tool calls in Tableau's activity and audit logs |
| Diagnose | Finds what is wrong in the activity record | Graph health suggestions on Cloud+ |
| Improve | The data team's answer goes into a draft | Business preferences change with no draft step. Revision history for workbooks and data sources |
| Verify | Evals: the draft against the live version | None for your own questions |
| Run a query | `sql_query`, one read-only statement on your database | `query-datasource`, a VizQL query on a published data source |
| Charts and dashboards | `render_chart` and `render_dashboard`, in the AI client | `get-view-image`, a PNG or SVG of a view saved in Tableau |
| Who can connect | Unlimited users, your customers' users included | People with an account on your Tableau site and API Access on the data source |
| Needed first | A data source: your database, or the sample data | Published data sources in Tableau Cloud or Tableau Server |
| Databases and warehouses | 7, plus PostHog | Published Tableau data sources, except cube data sources |
| Tools for the agent | Up to 17 | 48 in the docs. Site settings can exclude tools |
| How you pay | For each call, no seats | For each Creator, Explorer or Viewer, billed annually |
| Free plan | 3,000 calls a month | None. A free trial |

## When Tableau is stronger

- You want queries held to a governed model: `query-datasource` runs on a published data source, so its joins and calculated fields apply and the agent writes no SQL. Bonnard gives the agent the metric's SQL and does not rewrite the query.
- Your definitions already live in Tableau: the agent reads the descriptions, calculations and certifications your team has published, with each person's Tableau permissions.
- You want agents to read published dashboards: `get-view-image` and `get-view-data` return a view's image and data, after the filters that the agent passes in `viewFilters` ([get-view-data](https://tableau.github.io/tableau-mcp/docs/tools/views/get-view-data)).
- You want lineage and impact: on Cloud+, `query-knowledge-context` returns what feeds an asset and what a change to it affects.
- You run Tableau Server: the server is open source under [Apache-2.0](https://github.com/tableau/tableau-mcp/blob/main/LICENSE.txt) and you host it. OAuth needs Tableau Server 2025.3 or newer ([Tableau Server guide](https://tableau.github.io/tableau-mcp/docs/enterprise/tableau-server)).
- Your Tableau admins want agent help: when `ADMIN_TOOLS_ENABLED` is set, six prompts report on stale content, jobs and inactive licences. The three that make changes wait for a person to confirm ([prompts](https://tableau.github.io/tableau-mcp/docs/prompts)).

## Add each server to your AI client

**Bonnard.** Create a workspace, add Bonnard to your AI client, then add a data source ([quickstart](/docs/get-started/quickstart)). Use the install links for Claude, Cursor and ChatGPT in the console at `app.bonnard.ai`, or add this URL.

```text
https://app.bonnard.ai/mcp
```

**Tableau.** Hosted Tableau MCP serves every Tableau Cloud site. Site admins cannot disable it, but they can exclude its tools in site settings ([hosted Tableau MCP](https://tableau.github.io/tableau-mcp/docs/hosted-tableau-mcp)). In Claude, select **Customize > Connectors**, browse the connectors and add **Tableau Cloud**. In ChatGPT, install the **Tableau** plugin. Other clients use this URL ([client integrations](https://tableau.github.io/tableau-mcp/docs/hosted-tableau-mcp/client-integrations)).

```text
https://mcp.tableau.com
```

Sites with IP filtering add Salesforce's IP ranges to the allow list and use the URL of their Tableau Cloud pod ([IP filtering](https://tableau.github.io/tableau-mcp/docs/extras/ip-filtering)). Tableau Server customers host the open-source server themselves ([enterprise deployment](https://tableau.github.io/tableau-mcp/docs/enterprise)).

Verified 16 Nov 2026 against Tableau's [MCP docs](https://tableau.github.io/tableau-mcp/docs/intro), [hosted MCP docs](https://tableau.github.io/tableau-mcp/docs/hosted-tableau-mcp), [client integrations](https://tableau.github.io/tableau-mcp/docs/hosted-tableau-mcp/client-integrations), [tool pages](https://tableau.github.io/tableau-mcp/docs/category/tools), [get-datasource-metadata](https://tableau.github.io/tableau-mcp/docs/tools/data-qna/get-datasource-metadata), [query-datasource](https://tableau.github.io/tableau-mcp/docs/tools/data-qna/query-datasource), [search-content](https://tableau.github.io/tableau-mcp/docs/tools/content-exploration/search-content), [query-knowledge-context](https://tableau.github.io/tableau-mcp/docs/tools/knowledge/query-knowledge-context), [inspect-knowledge-context](https://tableau.github.io/tableau-mcp/docs/tools/knowledge/inspect-knowledge-context), [manage-knowledge-context](https://tableau.github.io/tableau-mcp/docs/tools/knowledge/manage-knowledge-context), [get-view-image](https://tableau.github.io/tableau-mcp/docs/tools/views/get-view-image), [get-view-data](https://tableau.github.io/tableau-mcp/docs/tools/views/get-view-data), [configuring Tableau](https://tableau.github.io/tableau-mcp/docs/configuration/tableau-config), [environment variables](https://tableau.github.io/tableau-mcp/docs/configuration/mcp-config/env-vars), [eval tests](https://tableau.github.io/tableau-mcp/docs/developers/eval-tests), [prompts](https://tableau.github.io/tableau-mcp/docs/prompts), [enterprise deployment](https://tableau.github.io/tableau-mcp/docs/enterprise), [Tableau Server guide](https://tableau.github.io/tableau-mcp/docs/enterprise/tableau-server), [IP filtering](https://tableau.github.io/tableau-mcp/docs/extras/ip-filtering), [licence](https://github.com/tableau/tableau-mcp/blob/main/LICENSE.txt), [pricing](https://www.tableau.com/pricing/cloud), [Tableau Knowledge](https://help.tableau.com/current/online/en-us/tableau_knowledge_about.htm), [revision history](https://help.tableau.com/current/pro/desktop/en-us/qs_revision_history.htm) and [VizQL Data Service limits](https://help.tableau.com/current/api/vizql-data-service/en-us/docs/vds_limitations.html).