# Bonnard vs Power BI: MCP server features compared

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

With Power BI (powerbi.microsoft.com), the agent writes DAX on one semantic model and runs as the person who signed in. With Bonnard (bonnard.ai), the agent writes read-only SQL on your database.

## 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.

Power BI also gives its context to the AI client. Microsoft documents two MCP servers for Power BI, and names Fabric IQ MCP as the one for business questions ([MCP servers overview](https://learn.microsoft.com/en-us/power-bi/developer/mcp/mcp-servers-overview)). This comparison is with Fabric IQ MCP, which is generally available. It has no tool that answers in natural language, so the agent in Claude or ChatGPT reads the model and writes the query ([Fabric IQ MCP](https://learn.microsoft.com/en-us/fabric/iq/connectors/fabric-iq-mcp)).

`GetSemanticModelSchema` returns the model's tables, columns, relationships and measures with their DAX expressions, plus the author's AI instructions and verified answers ([Fabric IQ skill](https://github.com/microsoft/skills-for-fabric/blob/main/skills/fabriciq/SKILL.md)). `GetReportMetadata` returns a report's pages, visuals and filters, and `ValueSearch` finds stored values. Authors write AI instructions and verified answers in Prep data for AI, a preview feature built for Copilot in Power BI ([Prep data for AI](https://learn.microsoft.com/en-us/power-bi/create-reports/copilot-prepare-data-ai)). A model holds up to 10,000 characters of [AI instructions](https://learn.microsoft.com/en-us/power-bi/create-reports/copilot-prepare-data-ai-instructions) and 250 [verified answers](https://learn.microsoft.com/en-us/power-bi/create-reports/copilot-prepare-data-ai-verified-answers).

The agent queries the semantic model with `ExecuteQuery` in [DAX](https://learn.microsoft.com/en-us/dax/dax-overview), Power BI's query language. It cannot run SQL on the source database and it cannot write. A measure that the agent calls by name is calculated as the model defines it. The agent can also write its own DAX calculation. Microsoft's skill tells it to prefer the model's measures and to treat AI instructions as mandatory, and without the skill the client decides.

### How wrong answers get found and fixed

Bonnard runs a loop of four steps on how agents use your data. Power BI has an audit log for the first step, and your team does the other three by hand.

[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.

Fabric's audit log lists two MCP operations, `ReadByPowerBIMCP` and `ModifiedByPowerBIMCP` ([operation list](https://learn.microsoft.com/en-us/fabric/admin/operation-list)). Confirm with Microsoft which Fabric IQ MCP calls they cover.

**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.]

Fabric IQ MCP has no suggestions from agent activity ([Fabric IQ MCP](https://learn.microsoft.com/en-us/fabric/iq/connectors/fabric-iq-mcp)). Your team works out what is wrong by hand.

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

On Power BI, an author changes the model. Fabric Git integration keeps semantic models in Azure DevOps or GitHub for review, and is in preview for semantic models ([Git integration](https://learn.microsoft.com/en-us/fabric/cicd/git-integration/intro-to-git-integration)). Verified answers do not work with Git integration or deployment pipelines.

**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.]

Fabric IQ MCP has no evals of agent answers. Authors test AI instructions and verified answers by hand in the Copilot pane, and a diagnostics download shows what Copilot used for an answer ([Prep data for AI](https://learn.microsoft.com/en-us/power-bi/create-reports/copilot-prepare-data-ai)). Those tests run on Copilot in Power BI and do not cover the agent in Claude or ChatGPT.

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.

On Power BI, your team finds a wrong answer when a person reports it, then fixes the model. Bonnard reads the agent's activity, suggests the fix and tests it before it goes live.

### 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.

Fabric IQ MCP starts from a Power BI semantic model or report that the person can read ([Fabric IQ MCP](https://learn.microsoft.com/en-us/fabric/iq/connectors/fabric-iq-mcp)). The reports and models do not need Fabric or Premium capacity. Your organisation's home region must run all Fabric workloads, so the server does not work in Power BI-only regions or sovereign clouds. The person needs no Build permission and no IDs, only the name or browser URL of the report or model. Answers are as current as the model's latest successful refresh. The context is as good as the measures, AI instructions and verified answers your team has put in the model.

### 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)).

Fabric IQ MCP signs each person in with a Microsoft Entra work or school account, through delegated OAuth 2.0 ([Fabric IQ MCP](https://learn.microsoft.com/en-us/fabric/iq/connectors/fabric-iq-mcp)). Service principals and app-only sign-in are not available. Power BI permissions, row-level security and object-level security apply to each query. Fabric IQ MCP has no route for your customers' users.

### 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)).

Microsoft's [Fabric IQ MCP docs](https://learn.microsoft.com/en-us/fabric/iq/connectors/fabric-iq-mcp) name no licence or plan for the server. The server runs each query as the person who signs in, so that person needs a licence that can read the content. Power BI Pro is USD 14 for each user a month, and Premium Per User is USD 24, both paid yearly ([Power BI pricing](https://www.microsoft.com/en-us/microsoft-fabric/features/power-bi/pricing)). To read content that other people publish, a person needs one of the two, and content in a Premium Per User workspace needs that licence ([licence types](https://learn.microsoft.com/en-us/power-bi/fundamentals/service-features-license-type)). The exception is content in a workspace on Fabric F64 or Premium capacity (compute that Microsoft bills by size), where a free licence is enough. Check the capacity price for your region on [Fabric pricing](https://azure.microsoft.com/en-us/pricing/details/microsoft-fabric/).

## Side by side

Table: Bonnard and Power BI MCP servers, verified 16 Nov 2026

| Feature | Bonnard | Power BI (Fabric IQ MCP) |
| --- | --- | --- |
| Who answers | The AI client the person uses, with Bonnard's context | The AI client the person uses, with the semantic model's context |
| Context the agent reads | Schema, descriptions, joins, metrics, business terms, searchable values | Tables, columns, relationships, measures, AI instructions, verified answers, report metadata |
| Context through MCP | `explore_schema`, for any AI client | `GetSemanticModelSchema` and `GetReportMetadata`, for any AI client that can sign in with Microsoft Entra |
| Metric definitions | The agent reads the SQL and writes the query | Measures in the model. A measure called by name is calculated as defined. The agent can also write its own DAX |
| Observe | Records each tool call, from any AI client | Audit log operations for MCP |
| Diagnose | Finds what is wrong in the activity record | By hand |
| Improve | The data team's answer goes into a draft | By hand. Git integration for semantic models (preview), not for verified answers |
| Verify | Evals: the draft against the live version | By hand in the Copilot pane. No evals of agent answers |
| Run a query | `sql_query`, one read-only statement on your database | `ExecuteQuery`, DAX on one semantic model, read-only |
| Charts and dashboards | `render_chart` and `render_dashboard`, in the AI client | No chart tool. `ExecuteQuery` returns rows |
| Who can connect | Unlimited users, your customers' users included | People with a Microsoft Entra work or school account and access to the model |
| Needed first | A data source: your database, or the sample data | A Power BI semantic model or report |
| Databases and warehouses | 7, plus PostHog | The data in Power BI semantic models |
| Tools for the agent | Up to 17 | 6, all read-only |
| How you pay | For each call, no seats | For each user (Pro or Premium Per User), or for capacity |
| Free plan | 3,000 calls a month | A free Fabric licence reads your own content, and content on F64 or Premium capacity |

## When Power BI is stronger

- Your measures already live in Power BI semantic models: the agent calls a measure by name and Power BI calculates it. Bonnard gives the agent the metric's SQL and does not rewrite the query.
- Agents must answer under each person's row-level and object-level security, with the permissions they already have in Power BI ([Fabric IQ MCP](https://learn.microsoft.com/en-us/fabric/iq/connectors/fabric-iq-mcp)).
- You want approved answers to common questions: `GetSemanticModelSchema` returns the model's verified answers, and Microsoft's skill tells the agent to use them first.
- You want agents to change the model: the Power BI Authoring MCP server creates, changes and deletes tables, measures, relationships and security roles ([Authoring server](https://learn.microsoft.com/en-us/power-bi/developer/mcp/power-bi-authoring-mcp)). Its local server, which does not run on macOS, works on Power BI Project files on disk, so you review its changes as a Git diff before you commit.
- You use Microsoft 365 Copilot, Copilot Studio or Microsoft Foundry, where Microsoft names Fabric IQ as the path for data answers ([MCP servers overview](https://learn.microsoft.com/en-us/power-bi/developer/mcp/mcp-servers-overview)).

## 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
```

**Power BI.** Microsoft's Fabric IQ MCP steps cover GitHub Copilot CLI: add the URL to `~/.copilot/mcp-config.json` (macOS or Linux), run `/mcp get FabricIQ` and sign in with Microsoft Entra ID. A client with no preregistered Entra application needs an app registration with the delegated `Item.Read.All` and `Item.Execute.All` permissions. Microsoft's Claude Desktop and ChatGPT steps cover only the hosted Authoring and Consumption servers ([external clients](https://learn.microsoft.com/en-us/power-bi/developer/mcp/remote-mcp-server-external-clients)). With private links, use `https://api.fabric.microsoft.com/v1/mcp/fabriciq` instead.

```text
https://fabriciq.svc.cloud.microsoft/v1/mcp/fabriciq
```

Verified 16 Nov 2026 against Microsoft's [MCP overview](https://learn.microsoft.com/en-us/power-bi/developer/mcp/mcp-servers-overview), [Fabric IQ MCP](https://learn.microsoft.com/en-us/fabric/iq/connectors/fabric-iq-mcp), [Fabric IQ skill](https://github.com/microsoft/skills-for-fabric/blob/main/skills/fabriciq/SKILL.md), [Prep data for AI](https://learn.microsoft.com/en-us/power-bi/create-reports/copilot-prepare-data-ai), [AI instructions](https://learn.microsoft.com/en-us/power-bi/create-reports/copilot-prepare-data-ai-instructions), [verified answers](https://learn.microsoft.com/en-us/power-bi/create-reports/copilot-prepare-data-ai-verified-answers), [operation list](https://learn.microsoft.com/en-us/fabric/admin/operation-list), [Git integration](https://learn.microsoft.com/en-us/fabric/cicd/git-integration/intro-to-git-integration), [Authoring server](https://learn.microsoft.com/en-us/power-bi/developer/mcp/power-bi-authoring-mcp), [external clients](https://learn.microsoft.com/en-us/power-bi/developer/mcp/remote-mcp-server-external-clients), [pricing](https://www.microsoft.com/en-us/microsoft-fabric/features/power-bi/pricing), [licence types](https://learn.microsoft.com/en-us/power-bi/fundamentals/service-features-license-type), [Fabric pricing](https://azure.microsoft.com/en-us/pricing/details/microsoft-fabric/) and [DAX](https://learn.microsoft.com/en-us/dax/dax-overview).