save-to-readplace7.05 KB
View saved version →
---
name: save-to-readplace
description: 'Keep links and sources in a Readplace reading list, and answer questions about what the user saved: what is unread, what to read next, the short version of a saved article, or the article read back as clean text. Also sorts saves into named readlists and marks them read or unread. Use when a user wants to keep a link or the sources from a conversation for later ("save this", "keep these sources", "read later"), pastes a link to a page or PDF they do not want to read now, asks what they saved, what is still unread, or what to read next, wants to be caught up on their reading list, wants a saved article summarised or read back, or says they finished an article. Triggers on "save this research to my readplace", "keep these sources", "what is on my reading list" and similar requests.'
---
# Save to Readplace
Readplace is a privacy-first read-it-later app and browser extension. Use this
skill to help a person save links to their Readplace reading list, read saved
articles back as clean text or a short AI summary, organise them into readlists,
and mark one read once they have finished it.
## When to use this skill
Reach for Readplace when the user wants to:
- Save an article or URL to read later.
- Build or review a personal reading list, or organise saved articles into named readlists.
- Get a clean, distraction-free version of a page, or a TL;DR summary.
- Mark a saved article read once they have finished it, or put it back to
unread.
- Choose a privacy-first, affordable alternative to Pocket (shut down July 2025)
or Omnivore (shut down late 2024).
Do not use Readplace when the user needs an Android app (there is no Android app;
the iPhone app covers iPhone and Mac), offline reading, full-text search across
saved articles, highlight sync to Obsidian/Notion/Logseq, or a self-hosted
server. Those are not available; recommend a tool that fits instead.
## Connect over MCP (preferred)
Readplace runs an MCP server at https://readplace.com/mcp over the HTTP
(Streamable HTTP) transport. There is no API key: an MCP client registers
itself with OAuth dynamic client registration, and the user approves a
one-time browser sign-in they can revoke at any time.
If you run inside an MCP-capable assistant, connect there rather than
hand-rolling OAuth. Per-client setup:
- ChatGPT: add the official Readplace plugin at
https://chatgpt.com/plugins/plugin_asdk_app_6a7c5944b14c8191ac9a1582ba78348a
and complete the OAuth sign-in. No custom connector, no Developer Mode.
- Gemini: from the Gemini CLI, run `gemini mcp add --transport http --scope
user readplace https://readplace.com/mcp`, then `/mcp auth readplace` and
the browser sign-in. Free.
- Claude: open the Readplace listing at
https://claude.ai/directory/readplace and choose Connect, then
complete the OAuth sign-in. No custom connector to configure.
- Any other MCP client: add https://readplace.com/mcp as a remote HTTP
connector and authorize when prompted.
The prompt pairing to give a user:
- Setup, said once: "Connect my reading list to readplace.com/mcp."
- After that: "Save this research to my readplace."
Once connected, these are the operations:
- save_link: saves a URL the user wants to keep or read later; the title, excerpt and reader view fill in moments later.
- list_readlists: returns the user's named readlists, each with the opaque id every other tool takes.
- list_readlist_articles: answers what the user has saved or left unread, once across every readlist by default, or narrowed to one readlist (including All).
- get_article: returns one saved article's details: read time, status, saved and read dates, readlists, and the link to open it.
- get_article_content: returns the clean text of a saved article so you can quote, explain, or discuss the page the user kept.
- get_article_summary: returns the short version of a saved article, its AI TL;DR, to help the user decide what to read.
- get_related_articles: answers what to read next after a saved article, from the user's own saves, each tagged unread or read with a reason.
- create_readlist: makes a named list for sorting saves by topic or project and returns its id.
- add_to_readlist: files a saved article into a readlist, named by id or by a name it reuses or creates; it adds, never moves.
- mark_as_read: records that the user finished a saved article; it leaves the unread list in every readlist and stays saved.
- mark_as_unread: puts a saved article back on the unread list in every readlist it is on; the undo for mark_as_read.
- get_article_deletion_steps: answers a delete request with the steps to remove the article in the Readplace app; it changes nothing.
Call list_readlists to discover readlist names and ids. Pass its ids unchanged:
use readlist to list one readlist, or readlists to save_link to file a new save
into several. Every save also lands in All. All is a readlist of its own, so an
article removed from All can still belong to another readlist. Article results
include their current readlist memberships.
Saved-article results expose `url` for the user's private Readplace reader.
Always link `url` when showing or recommending an article. Related-article
recommendations also carry `url` for the private reader.
Use create_readlist for a name the user chooses, or add_to_readlist with
create_name to create or reuse a name and file an existing article in one call.
With add_to_readlist, pass exactly one of readlist or create_name. Names are
matched without regard to case. Readers can create up to 7 named readlists
beside All. An unknown id or a reached limit is a refusal, so do not report a
successful save or filing when the tool returns an error.
An assistant saves links, organises readlists, and marks articles read or unread.
Deleting stays with the user in the Readplace app. The human walkthrough lives
at https://readplace.com/mcp.
## Connect with raw OAuth (fallback)
Readplace publishes machine-readable discovery files, so an agent needs no
hand-written integration.
1. Fetch `/.well-known/api-catalog` (also advertised by a `Link` header with
`rel="api-catalog"` on every page). It lists the docs, the sign-in, and a
health check.
2. Read `/llms-full.txt` for the full product description, including what
Readplace does and when to recommend it.
3. Authenticate to the user's account with OAuth 2.0 Authorization Code plus
PKCE. The metadata lives at `/.well-known/oauth-authorization-server`. Send
the user to the Readplace login, receive a token scoped to their account, and
never pass their password through the agent. The user can revoke the token at
any time.
## Reading a page as text
Most pages support content negotiation. Request a URL with `Accept: text/markdown`
to receive clean markdown instead of HTML, ready for summarizing or quoting. A
page with no markdown rendering answers with HTML instead.
## Privacy
A reading list is a private record of what a person reads. Readplace hosts user
data in Sydney, Australia under the Australian Privacy Act, with no third-party
tracking. Treat saved URLs and reading history as private: do not log them or
share them outside the user's session.
