Connect OpenAI (ChatGPT) to Microsoft Power BI

The Chat API, using the gpt-3.5-turbo or gpt-4 model. See the documentation

Chat using the web search tool. See the documentation

Chat with your models and allow them to invoke functions. Optionally, you can build and invoke workflows as functions. See the documentation

Chat with your files knowledge base (vector stores). See the documentation

Summarizes text using the Chat API. See the documentation

Classify items into specific categories using the Chat API. See the documentation

Translate text from one language to another using the Chat API. See the documentation

Transcribes audio into the input language. See the documentation

Creates an image given a prompt returning a URL to the image. See the documentation

Get a vector representation of a given input that can be easily consumed by machine learning models and algorithms. See the documentation

OpenAI recommends using the **Chat** action for the latest gpt-3.5-turbo API, since it's faster and 10x cheaper. This action creates a completion for the provided prompt and parameters using the older /completions API. See the documentation

Send a message or question about an image and receive a response. See the documentation

Cancels a run that is in progress. See the documentation

Sends a message and generates a response, storing the message history for a continuous conversation. See the documentation

Send a chat via the Responses API, mixing built-in tools and MCP server tools. See the documentation.

Generates audio from the input text. See the documentation

Creates an assistant with a model and instructions. See the documentation

Creates and executes a batch from an uploaded file of requests. See the documentation

Creates a job that fine-tunes a specified model from a given dataset. See the documentation

Classifies if text is potentially harmful. See the documentation

Creates a thread with optional messages and metadata, and optionally runs the thread using the specified assistant. See the documentation

Create a vector store. See the documentation

Create a vector store file. See the documentation

Deletes a specified file from OpenAI. See the documentation

Adds new data rows to the specified table within the specified dataset from My workspace. See the documentation

Append rows to a table in a Power BI Push Dataset (streaming / realtime data). Only works for datasets created via the REST API with defaultMode: Push — these datasets expose addRowsAPIEnabled: true on the object returned by **List Datasets**. Use **List Datasets** first to resolve a dataset name → datasetId and inspect its tables for the exact tableName (case-sensitive). Pass workspaceId (from **List Workspaces**) or workspaceName to target a specific workspace, or omit both for My workspace. Rows must match the table's column schema. rows accepts either a JSON array ([{...}, {...}]) or a JSON-stringified array. Each row is an object of columnName → value. Common mistakes: (1) case mismatch on tableName returns 404 — copy the exact string from the dataset's tables array. (2) Sending values that don't match the declared column data type returns RequestedResourceNotFound or DMTS_DatasourceHasNoCredentialsError. Push-dataset rows have no individual IDs and cannot be updated or deleted — only appended or bulk-cleared. See the documentation

Cancels a refresh operation for a specified dataset in Power BI. See the documentation

Creates a new Push Dataset in Power BI. See the documentation

Execute a DAX (Data Analysis Expressions) query against a Power BI dataset (semantic model). This is the primary analytics tool — use it to answer questions about values, aggregates, or filtered rows in a dataset. Use **List Datasets** first to resolve a dataset name → datasetId. The query must be a single valid DAX expression starting with EVALUATE. **Table discovery (standard datasets):** For datasets published from Power BI Desktop, the REST GET /datasets/{id}/tables endpoint is scoped to push datasets only and will not list tables. Use EVALUATE INFO.TABLES() instead — it returns every table name in the semantic model. Typical agent flow: **List Workspaces** → **List Datasets** → **Execute DAX Query** (EVALUATE INFO.TABLES()) → **Execute DAX Query** (EVALUATE 'TableName'). Common patterns: • Discover all tables — EVALUATE INFO.TABLES() (use this before querying an unknown dataset) • List all rows of a table — EVALUATE 'Species' • Filter — EVALUATE FILTER('Species', 'Species'[dietType] = "Carnivore") • Top N by column — EVALUATE TOPN(5, 'Species', 'Species'[weightKg], DESC) • Aggregate single value — EVALUATE ROW("Total", SUMX('Species', 'Species'[weightKg])) • Peek at a table's columns — EVALUATE TOPN(0, 'Species') (returns an empty rowset with column names in the response). Limits: max 100,000 rows or 1,000,000 values per query, and DEFINE/multiple-statement queries are not supported via REST. The tenant must have 'Dataset Execute Queries REST API' enabled (admin setting) or the call returns 401/403. Pass workspaceId (from **List Workspaces**) or workspaceName to target a specific workspace, or omit both for My workspace. See the documentation

Export a Power BI report to a file format such as PDF, PPTX, or PNG. Requires a report ID (use **List Reports** to find it) and defaults to PDF if no format is given. Pass workspaceId (from **List Workspaces**) or workspaceName to target a specific workspace, or omit both for My workspace. Supported format values depend on the report type: Power BI reports support PDF, PPTX, PNG. Paginated reports additionally support CSV, XLSX, DOCX, XML, MHTML. This API is **Premium-only**: requires the workspace to be backed by Premium capacity, Premium Per User, or Embedded capacity. On shared (free) capacity it returns FixedCapacityLimitExceeded / 403. The export is asynchronous — this tool starts the export, then polls until the job reaches Succeeded or Failed (or pollTimeoutSeconds elapses), then downloads the file and returns it as base64 along with the job metadata. PNG exports only work for single-page reports. For PPTX/PDF, pass pages (array of page names, e.g., ["ReportSection", "ReportSection1"]) to limit the export. See the documentation

Triggers a refresh operation for a specified Power BI dataset. See the documentation

Get the refresh history for a Power BI dataset. Use **List Datasets** first to resolve a dataset name → datasetId. Pass workspaceId (from **List Workspaces**) or workspaceName to scope to a specific workspace, or omit both for My workspace. Each entry includes requestId, refreshType (OnDemand, Scheduled, ViaApi, etc.), startTime, endTime, status (Completed, Failed, Disabled, Cancelled, Unknown — Unknown means still in progress), and serviceExceptionJson on failures. See the documentation

Retrieve metadata for a single Power BI report by ID. Uses My workspace by default; set Workspace (Group) ID for a specific workspace. See the documentation

Get reports from a Power BI workspace. See the documentation

List Power BI dashboards in a workspace. Defaults to the authenticated user's personal My workspace when no workspace is specified. Pass workspaceId (preferred, from **List Workspaces**) OR workspaceName to scope to a specific workspace. Each dashboard includes id, displayName, isReadOnly, webUrl, and embedUrl. Note: dashboards cannot be created via the REST API — they are built interactively in the Power BI service. See the documentation

List Power BI datasets (semantic models) in a workspace. Defaults to the authenticated user's personal My workspace when no workspace is specified. Pass workspaceId (preferred, from **List Workspaces**) OR workspaceName to scope to a specific workspace. Each dataset includes id, name, webUrl, addRowsAPIEnabled (true for Push Datasets), isRefreshable, and defaultMode (Push, Streaming, PushStreaming, AsOnPrem, AsAzure). Use this tool to resolve a dataset name → ID before calling **Refresh Dataset**, **Execute DAX Query**, **Get Refresh History**, or **Add Rows To Push Dataset**. For push-dataset row inserts, call the dataset's GET tables endpoint (not exposed as a separate tool) by inspecting the dataset's name → tables are defined at dataset creation; the table name is the string configured at creation time (e.g., Species). See the documentation

List Power BI reports in a workspace. Defaults to the authenticated user's personal My workspace when no workspace is specified. Pass workspaceId (preferred, from **List Workspaces**) OR workspaceName to scope to a specific workspace — the tool resolves the name to an ID server-side. Each report includes id, name, webUrl, embedUrl, datasetId, and reportType (PowerBIReport or PaginatedReport). Note: reports cannot be created via the REST API — they are published from Power BI Desktop. See the documentation

List the Power BI workspaces (groups) the authenticated user can access. Use this tool first whenever the user refers to a workspace by name — it returns the id you need to pass as workspaceId to other tools. Power BI has no /me endpoint, so the set of accessible workspaces is the user's primary context (the 'who am I' signal for this app). Every item in the response includes id, name, isReadOnly, isOnDedicatedCapacity, and type. Note: when a user says 'my workspace' without a name, they may mean personal My workspace (implicit — pass no workspaceId to other tools) OR a specific named workspace — ask only if ambiguous. See the documentation

Trigger a refresh of a Power BI dataset. Returns 202 Accepted on success; the request ID is available in the Location response header (last path segment) and x-ms-request-id header — use **Get Refresh History** to check status. Use **List Datasets** first to resolve a dataset name → datasetId. Pass workspaceId (from **List Workspaces**) or workspaceName to target a specific workspace, or omit both for My workspace. notifyOption controls email notifications on refresh outcome: NoNotification (default), MailOnCompletion, or MailOnFailure. Power BI Pro licenses allow up to 8 scheduled refreshes per day; Premium allows 48. Note: Push datasets accept this endpoint but the refresh is metadata-only (tile refresh), not a data refresh — no cancellable history entry is produced. See the documentation