Readwise/AI/Report/26/01/AI control of Readwise - Options Jan 2026
Properties
Tools to Connect AI Agents with Readwise Reader & Highlights
Command-Line Interfaces (CLI)
- Official CLI: There is currently no official Readwise CLI tool. However, open-source alternatives exist for command-line interaction with Readwise Reader.
- Readwise Reader CLI (rw-cli) – An open-source Python CLI by Scarvy that interacts with the Readwise Reader API. It supports adding new items (documents) to your Reader library and listing documents with various filters [1]. For example, you can use
rw-cli add <URL>to save an article, orrw-cli list --location later --category articleto list saved articles (with options to filter by location like new, later, archive, or by category and date) - This tool is actively maintained (over 130 commits) and is community-supported.
- Note: Other community scripts exist, such as a LaunchBar action "Reade" for macOS [3]. This tool is the primary CLI for Readwise Reader.
Model Context Protocol (MCP) Servers
- Official MCP Server: Readwise provides an official MCP server (@readwise/readwise-mcp on NPM) which is maintained by the Readwise team. This server allows AI clients (like Claude or Cursor) to search and retrieve highlights from your Readwise account [4]. It focuses on the main Readwise highlights (e.g. Kindle/book highlights) rather than full Reader documents, and is regularly updated (v0.0.7 as of late 2025).
- Open-Source MCP Servers: Several community-driven MCP servers extend Readwise integration, enabling AI agents to read/write Reader documents and highlights. Notable examples include:
- Readwise Reader MCP (edricgsh) – A popular open-source TypeScript MCP server that exposes the entire Readwise Reader API to AI agents [5]. It gives your AI full control over your Reader library, supporting full CRUD on documents (create/save new articles, list/filter documents, update metadata like tags/folder, and delete documents) as well as tag listing and powerful regex search across content [5]. This server is built with the official Claude SDK and is actively maintained; it had ~3.6k downloads by Oct 2025 [5], indicating healthy usage.
- Readwise MCP Enhanced (arnaldo-delisio) – An open-source "unified" MCP server combining both Reader documents and Readwise highlights in one tool [7]. Written in TypeScript, it integrates the Reader API (v3) for document management and the original Readwise API (v2) for highlights [7]. This server provides comprehensive features (saving/searching documents, accessing all highlights, listing books and tags, etc.) with various optimizations (e.g. reduced token usage, content pagination) [7]. It is fairly active (40+ stars and ongoing development) and is compatible with any MCP-enabled AI client (Claude, Continue IDE, Raycast, etc.).
- Readwise MCP HTTP Server – A Node.js MCP server (by CaseyRo) that exposes Readwise functions over standard HTTP endpoints [9]. This allows AI tools (or any HTTP client) to query and retrieve highlights or Reader documents via HTTP requests, which can simplify integration. It's an open-source project (TypeScript) focused on providing MCP-over-HTTP for easier debugging and integration [9]. (This project is smaller in adoption, but useful if your AI platform prefers HTTP calls over the MCP protocol).
- Remote Readwise MCP (Python) – An unofficial Python-based MCP server for Readwise [10]. It offers similar capabilities to the above (primarily Reader document access) but implemented in Python, which may appeal to those in Python ecosystems. This server (e.g. xinthink/reader-mcp-server-py) supports listing and retrieving Reader documents [10]. Its community usage is more niche compared to the TypeScript solutions, but it remains a viable open-source option.
- Maintenance: All the above MCP servers are open-source. The official MCP server is maintained by Readwise (commits as recent as Sep 2025), while the community servers are maintained by individual developers. As of late 2025, edricgsh's and Arnaldo's servers show active development and community engagement [5] [7].
SDKs and Libraries
- For developers looking to integrate Readwise into projects (or AI agent code) directly, there are several official or unofficial SDKs and API wrappers:
- PyReadwise (Python) – An unofficial Python package for the Readwise API [12]. It provides convenient methods to access Readwise highlights and associated data through the official API. With PyReadwise, a Python-based agent can fetch highlights, create new highlights, or sync data without manually calling HTTP endpoints. (This library targets the original Readwise v2 API, focusing on books/articles highlights management.) [12]
- Readwise Reader API Client (Python) – A community-maintained Python library (e.g. floscha/readwise-api) that wraps the Readwise Reader API [12]. It enables Python applications or agents to programmatically add documents to Reader and list/search documents using high-level functions instead of raw HTTP calls. This SDK corresponds to Readwise's Reader endpoints (v3), making it easier to integrate "save to Reader" or "fetch my articles" features in Python. (It's a smaller project with ~7⭐, but useful for quick integration.) [12]
- Readwise Reader API JS/TS SDK – An unofficial JavaScript/TypeScript client (by Scarvy) for the Readwise Reader API [12]. This package simplifies using Reader endpoints in Node.js or browser contexts. Using this, a JS-based AI agent (or tools like a VS Code extension) can add documents, list items, etc., with TypeScript types and convenience methods. It's open-source (≈11⭐) and often updated alongside the CLI tool by the same author. [12]
- Readwise Rust SDK – A Rust crate (readwise) that wraps the Readwise API [12]. Geared towards systems programming or Rust-based backends, this library allows fetching and uploading highlights via the official API. It was created to enable integration of Readwise into Rust projects (e.g. knowledge management tools). Maintenance activity is moderate (8⭐), and it may primarily support the highlights (v2) endpoints as of now [12].
- Other languages: Additional unofficial API wrappers exist, such as readwise-go (Go library for the Readwise API) and readwise-ruby (Ruby gem for highlights) [12]. These are community projects that similarly cover highlight retrieval and syncing. While not in Rust/Python/TS, they indicate the ecosystem's breadth and may be useful if you use those languages.
- Maintenance: These SDKs are third-party projects. Many are open-source on GitHub with small communities. Before using one, check its last update – e.g., PyReadwise and the TS client have recent commits in 2025, whereas some others might be older. In absence of official SDKs from Readwise, these libraries provide a starting point for integration.
- For developers looking to integrate Readwise into projects (or AI agent code) directly, there are several official or unofficial SDKs and API wrappers:
Direct Readwise API Access
- All the above tools ultimately rely on Readwise's official APIs. You can integrate an AI agent directly with these HTTP APIs if no suitable tool exists for your use case:
- Readwise Highlights API (v2) – The primary REST API to Readwise (the original service) supports creating, fetching, and updating highlights [16]. This includes endpoints to get all highlights or filter by book/article, create new highlights (e.g. via POST /api/v2/highlights/), update highlight text or tags, etc. An AI agent could call these endpoints directly using HTTPS. For example, to add a highlight, you would use the Highlight CREATE endpoint with an API token [16]. The API returns structured JSON (books and highlights), and has a generous rate limit (~240 req/min for most calls) [16]. This direct approach gives full control, but requires handling HTTP requests and auth tokens in the agent's code.
- Readwise Reader API (v3) – A separate set of endpoints dedicated to the Reader service (read-it-later documents). The Reader API allows you to save new documents to Reader and list/search your documents programmatically [20]. For instance, an AI agent can POST to https://readwise.io/api/v3/save/ with a URL (and optional HTML content or tags) to add an article to your Reader library [20]. It can also GET https://readwise.io/api/v3/list/ with query parameters to retrieve documents – filtering by location (new, later, archive, feed), category (article, pdf, email, etc.), tags, or updatedAfter a given date [20]. This enables assembling lists of recently read or highlighted items by specifying a date filter (for example, all items updated in the last week) [20]. Tags can be listed via a GET /api/v3/tags/ endpoint as well [20]. Using the API directly is fully supported (the documentation is available on Readwise's site), though it requires more manual effort.
- Optional Features & Considerations: Some features (like managing RSS feeds in Reader) are not yet exposed via the public API. According to Readwise, the Reader API is currently limited to documents and will gain more endpoints over time [20]. This means subscribing/unsubscribing to feeds or creating feed folders must be done through the Reader app for now (no official API endpoints as of 2026). Likewise, downloading full article text can be achieved indirectly: the Reader API's list can return an html_content or a raw_source_url for each document when requested [20], which an agent can use to fetch the full text. Tools like the Readwise MCP servers also handle content extraction (e.g. converting HTML to clean text for the AI) [25] [26]. Finally, listing tags and basic metadata is supported (via the tags endpoint or MCP server functions) in most integrations, even if feed/category CRUD is absent.
- All the above tools ultimately rely on Readwise's official APIs. You can integrate an AI agent directly with these HTTP APIs if no suitable tool exists for your use case:
Summary
- In summary, to enable an AI agent (Cursor, Claude's code assistant, Codex, etc.) to interface with Readwise, you can choose from a range of solutions: a convenient CLI tool [1], specialized MCP servers for real-time AI context [4] [5], language-specific SDKs [12], or direct use of Readwise's REST APIs [16]. Official offerings exist for the MCP and the APIs, whereas community open-source projects fill in the gaps with additional capabilities (especially for Reader). Depending on your workflow and preference (command-line vs. programmatic, language ecosystem, real-time AI context needs), you can pick the appropriate tool. All options support the core features — saving to Reader, retrieving highlights, and assembling reading lists — while more advanced operations (feeds and category management) will become available as the Readwise API expands [20].
Footnotes