The problem with this approach is that for better/worse, Paperpile is very, very slow to get new features (and/or, they are spread across enough platforms that no specific platform gets very many few features). Fortunately for PP, a good chunk of your audience know how to program (i.e., those in sciences, computer science, etc) and would likely build things on top of the public API to address shortcomings.
3 Likes
I think you underestimate the value add of plugins + API features. Particularly with the rise of LLMs.
4 Likes
Any suggestions on how to get highlight summaries out of PaperPile and into Roam Research, as a public API doesnāt seem to be a priority for the roadmap (in this day and age??)
It may be possible in upcoming time while till now it is not apper. In future it may be happened.
Thanks
I would like to add a +1 for the addition of a Public API. The possibilities seem almost mind-blowing, AI/LLM analysis, Obsidian etc. integration, creating value-add services etc. Iām doing quite a lot of this just now, but there is just this manual bit where Paperpile comes in. I tried using Zotero and it opens up a lot of possibility but itās just not quite worth loosing the great PP functionality ā especially with the new Beta site and full-text search.
To PP product dev team: would it help if a few of us got together and drew up a rough outline of what a MVP API would look like with some use cases? It seems like thereās plenty of interest, and perhaps some clarification of exactly what we want might be useful rather than āGIVE US APIā 
4 Likes
Why is a public API not a high priority given high interest of your customers throughout the years?
3 Likes
yes! building a core, extensible platform like Obsidian.md has to be the #1 priority after getting the new app* released. Allowing power users / developers to build creative and/or niche add-ons will boost development velocity many times over compared to what Paperpile dev team alone is capable of.
Regarding @vicenteās comment:
it is a matter of user demand
if you build it, they will come. Value Propositions in Business Ecosystems - Products, Marketplaces, Extensions - Boundaryless
*hopefully itās been developed with this in mind already so there isnāt a ton of refactoring.
4 Likes
I just want to add my voice to the chorus. With Obisidian & LLMs increasingly becoming a staple of the academic workflow (not to mention a myriad other possibilities that an API opens up) I feel like this should be a top priority for your dev team. You must be using an API for your phone apps, why not just expose some of that to the user side?
The only thing keeping me from moving to Zotero has been the pain of transferring my annotated pdfs along with their metadata.
If anyone has a solution for this, please let me know.
4 Likes
What all my excuses, I am new to this platform and I am inserting myself into this thread.
- How do PaperPile compile your library? EndNote use an arbitrary reference number, Zotero an arbritary alphanumeric code, that is, there is a lot is arbitrary. In medicine, one could use PubMed ID pmid to indentify in an absolute manner a reference, that is, it is an ID independent of any personnal library.
- Is there an API that I could use for interfacing with, say, Python. I would provide Python with a list of, say, DOIs, Python would interrogate Paperpile, and Paperpile would provide the needed, say, in-text citations.
- How does PaperPile format the in-text and bibliography entries? EndNote can format and unformat in MS Word. Zotero says that the formatting is not reversible, and advise to safeguard the original manuscript. What about PaperPile and its MS Word addin? Is it reversible?
Richard, attempting to write a to be shared medical encyclopedia entitle the Internist Browser
2 Likes
I wonder if the API have been released or will be released. I guess this will really help Paper pile in this wave of AI. By linking personal Paperpile database to something like Notebook LLM can be really useful.
2 Likes
Just to add my voice, this would be a killer feature!
1 Like
So this is now 10 years old - where are we on the roadmap with this one? 
Quick update on this well-aged thread: the team has broken ground on building a public REST API, which we hope to get into a few usersā hands within the next several weeks for early testing. Beyond the API (which will stand on its own) we plan to build a layer of infrastructure to enable powerful integrations with agent-based tools like Claude Code. See this post for a bit more detail.
Something Iāll repeat from that post: Iād be eager to hear feedback and concrete use cases youād like us to pay attention to when developing these APIs. We do our best to use good judgment and target a flexible but not overwhelming API, but itās always better to hear from potential API users about their motivating use cases and priorities. Donāt hesitate to get in touch if you have thoughts!
Itās an exciting moment for the Paperpile team ā weāre (finally) in a place where enough of the critical-path product and infrastructure improvements are either completed, or far enough along, that we can invest strategically in opening up Paperpile as a platform to developers. And itās especially timely, as coding agents are greatly expanding everyoneās ability to build new tools to facilitate their work. Iām eager to see what new workflows and tools emerge that we wouldnāt have been able to dream up on our own!
3 Likes
Exciting!
This is exciting news, @greg! I was seriously looking at switching to Zotero so that I could get agentic integration, but now Iāll wait and see how this goes.
Key use cases for me include:
- Allowing an agent to search across papers and summarize them, looking for common themes, answers to questions, etc.
- Having an agent organize papers into collections or tag them in various ways to form coherent collections
- Having an agent search various sources for additional work related to certain papers, and add them automatically to Paperpile.
If thereās an opportunity to get early access to the MCP features / beta test, Iād love to sign up and get started on some of these.
1 Like
Iād love to participate in early testing; my own personal use case would be to use Paperpileās elegant āCAPTCHA, user input neededā functionality (if possible!) as I try and code paper retrieval processes for some studies Iām working on.
Thanks Tristan. Weāll keep you posted as soon as thereās something to start working with.
I think youāll be happy to see what weāre working onā¦ your use cases fit our goals almost to the letter. 
Out of curiosity: if thereās a particular agent harness thatās a must-have for you right now, Iād love to hear about it (feel free to email greg@paperpile.com if youād prefer a side channel). Weāll target all platforms eventually, but we need to focus our efforts early on.
Iād also be curious to learn what other MCP servers, skills / plugins researchers are actively working with and finding useful.
my own personal use case would be to use Paperpileās elegant āCAPTCHA, user input neededā functionality (if possible!) as I try and code paper retrieval processes for some studies Iām working on.
You wouldnāt believe the amount of time our team has spent worrying about CAPTCHAs. I wouldnāt have guessed this would be the state of things in a world where bandwidth is cheap and open access is on the rise. Yet here we are
Iām not sure weāll have this use case nailed down perfectly at first, but we have some ideas in the works. Iāll reach out if we have ideas in this direction weād like to get feedback on.
Personally Iām using Claude almost exclusively - through Claude Code, Desktop and the Web app. An MCP server that integrates with these would be the best fit for me.
I mostly use the connectors for Google workspace (Gmail, calendar, drive) as well as some other productivity apps. I use an MCP for code documentation (context7) and I use a variety of different skills depending on the task: some Iāve sourced from open collections and some Iāve created myself by distilling e-textbooks that I trust.
Excited to hear, that there is work going on regarding an API and MCP!
I am currently figuring out how to integrate agents into my workflow, so I would be happy to test any MCP features as well.
My current use case would be to export paper metadata and notes to Obsidian. Something I would love would be the possibility to link from Obsidian to papers in Paperpile.