Get Answers From Your Highlights (using ChatGPT)
Based on Readwise's video on YouTube. If you like this content, support the original creators by watching, liking and subscribing to their content.
Import highlights into Readwise first so there’s a searchable highlight database for ChatGPT to query.
Briefing
Readwise MCP turns ChatGPT into a personal research assistant that can answer questions using a user’s own reading highlights—so the model can retrieve quotes, refresh complex ideas, and synthesize themes across books instead of relying on general web search. The workflow starts with getting highlights into Readwise, then installing the Readwise MCP connector inside ChatGPT, and finally querying a highlight database that ChatGPT can search automatically.
The first prerequisite is importing highlights into Readwise so there’s something for ChatGPT to reference. Users can connect and sync from the Readwise dashboard by going to “Connect and Sync” and choosing “import.” The import page supports more than 30 sources, including Kindle and Pocket, plus Readwise’s all-in-one reading app called Reader (available via free trial and subscription). Readwise also supports importing highlights from physical books through a mobile “physical book import” feature. Once highlights are in place, the next step is enabling the MCP connector inside ChatGPT.
Access to MCP is gated by ChatGPT plan level: only users on pro team or enterprise plans can use the MCP functionality. Free and Plus users must upgrade to access it. Setup begins in ChatGPT settings (bottom-left), under “connectors,” where users add a new connector named for clarity (e.g., “Readwise” or “Readwise MCP”). The connector requires the Readwise MCP server URL from Readwise documentation, and authentication uses OAuth. Users should select “Trust this application” and then authorize the connection on the Readwise website. OpenAI is explicitly noted as not using Readwise data to train its models.
After installation, users enable “Connected Apps” and select Readwise as the source. Prompts don’t need to include the word “highlight” because ChatGPT defaults to searching the Readwise highlight database. One key capability is semantic (fuzzy) search: users can describe a quote they remember only roughly, and ChatGPT searches a highlight vector database to locate the closest match. In the example, ChatGPT finds an anecdote from Byun-Chul Han’s The Crisis of Narration and returns a fuller description first; a follow-up request retrieves the exact highlight for direct quotation.
Beyond single retrieval, the MCP supports synthesis across multiple highlights. Users can ask ChatGPT to refresh a complex concept from a specific book using the highlights they saved, then follow up with related questions such as solutions proposed by the author. It can also aggregate across documents—for instance, outlining common mistakes early product managers make—while optionally formatting responses for readability. The practical payoff is faster recall for essays, reviews, and sharing quotes without manually flipping through highlight lists, plus the ability to generate structured takeaways grounded in personal notes.
Cornell Notes
Readwise MCP lets ChatGPT answer questions using a user’s own Readwise highlights, turning saved reading into a searchable knowledge base. The setup requires importing highlights into Readwise from sources like Kindle, Pocket, Reader, and even physical books via mobile import. After installing the Readwise MCP connector in ChatGPT (available only for pro team or enterprise plans), users authenticate via OAuth and enable Connected Apps. ChatGPT can then perform semantic search to find a specific highlight from a rough description, return the exact quote, and link back to the original source context. It also synthesizes larger takeaways by combining highlights from one book or across multiple documents.
What must be done before ChatGPT can answer questions from personal reading highlights?
Why can’t every ChatGPT user immediately use Readwise MCP?
How does semantic search work when trying to find a quote that isn’t remembered exactly?
How can a user view a found highlight in context with other highlights from the same document?
What’s the difference between retrieving one highlight and synthesizing broader answers?
Review Questions
- What import sources does Readwise support for getting highlights into the system before MCP can search them?
- How does semantic search help when you can’t remember the exact wording of a quote?
- What kinds of questions are best suited to single-highlight retrieval versus cross-document synthesis?
Key Points
- 1
Import highlights into Readwise first so there’s a searchable highlight database for ChatGPT to query.
- 2
Readwise MCP setup in ChatGPT requires pro team or enterprise access; free and Plus users must upgrade.
- 3
Install the Readwise MCP connector via ChatGPT settings → connectors, using the Readwise MCP server URL and OAuth authentication.
- 4
Enable Connected Apps and select Readwise as the source; prompts can omit the word “highlight” because ChatGPT defaults to the highlight database.
- 5
Use semantic search by describing a quote or idea roughly; ChatGPT searches a vector database to find the closest highlight.
- 6
Ask follow-up questions to extract exact quotes or to synthesize explanations and solutions based on the highlights saved.
- 7
Leverage hyperlinks from ChatGPT results to view a highlight in the context of other highlights from the same document.