# March 6: Support for MCP Server, Mistral OCR, retrieveSources, GPT-4.5, Sonnet 3.7, bug fixes

### New Features

* :bulb: Graphlit now supports the Anthropic MCP (Model Context Protocol) standard, through our new open source MCP server.  [Read more in our blog](https://www.graphlit.com/blog/graphlit-mcp-server).
* Graphlit now supports the `retrieveSources`mutation in the API, which provides reranked retrieval of content sources, for use in agents and MCP clients, outside of RAG conversations. We have also added support for filtering retrieved sources by feed(s) and collection(s).
* Graphlit now supports the [Mistral OCR model](https://docs.mistral.ai/capabilities/document/) for document preparation, with the preparation service enum of `MISTRAL_DOCUMENT`.  When selecting this document preparation service, it will use the `mistral-ocr-latest`model with PDFs. (Note: There are known issues with the model extracting embedded images without parsing text from them.)
* We have added support for OpenAI GPT-4.5 model with the model enum `GPT45_PREVIEW_128K`.
* We have added support for the Anthropic Claude 3.7 model with the model enum `CLAUDE_3_7_SONNET`, including the [thinking token budget.](https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking)
* When uploading files via the Graphlit Tus server, we now support the assignment of `workflowId`and `collectionId`, so that the ingested file can automatically be assigned to a collection, and have the specified workflow applied.
* We have added the `linearProjects`query to the API, for enumerating Linear projects in your authenticated account.

### Bugs Fixed

* GPLA-3912: Can't upload same filename twice via Uppy
* GPLA-3892: Failed to process table with Azure Doc Intelligence
* GPLA-3869: Failed to get Microsoft email delta
* GPLA-3947: Failed to parse hyperlinks from Word doc
* GPLA-3878: PDF extraction can be more detailed on charts
* GPLA-3830: Failed to get Google Drive URI (invalid chars in filename)
* GPLA-3908: Not adding line break with Sonnet markdown extraction


---

# Agent Instructions: Querying This Documentation

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

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://changelog.graphlit.dev/march-2025/readme-5.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
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.