Changelog

You put together a collection for a project, and your Copilot agent still searched the whole library. Asked “What’s new?”, it answered by publication year, not by what you added most recently. It can do both now.

• Find collections and search inside them: Asked “Is there a collection on accreditation?”, the assistant names the matching collections with their description and number of publications. It shows what is in a collection and, if you ask, searches only inside it. It only sees collections you can see in Librario as well.
• Show new additions: Asked “What was added last week?”, it lists the most recently added entries, regardless of publication year. A 2019 standard you catalogued yesterday is at the top. It also tells you when an entry was added to the library.

Your AI assistant also knows more about your holdings:

• Standard number and validity: It names a standard’s number, such as “DIN 1045-1:2008-08”, and your internal shelfmarks. Asked “Is DIN 14230 still valid?”, it answers from your holdings: whether the standard is valid or withdrawn, which edition it replaced and which replaces it. It goes by the status you recorded in Librario; Librario does not check it against the DIN catalogue.
• Exact publication date: It gives the publication date as precisely as you recorded it, with month or day. Asked “When was this published?”, it answers “January 2020”. Where you recorded only the year, it gives only the year.

More about the connection is in the MCP documentation.

Your team builds its own agents in Microsoft Copilot Studio, and your specialist literature lives in Librario. Such an agent now researches your library directly: it finds publications on a question, gives a standard’s number and whether your records mark it as still valid, and offers attached files for download.

By default everyone signs in with their own Librario user account. The agent then sees only what that person may see in Librario, and your IT can lock this setting for every agent. You set up the connection inside the agent: Copilot Studio registers itself with Librario, and nobody has to create credentials in Librario first. You need a Copilot Studio license from Microsoft and a Librario plan with access for AI assistants, but no add-on.

The Copilot Studio connection is in beta. The steps are in the Copilot Studio guide. Tell us how it works for you through the contact page.

Scripts and integrations that create publications usually name authors, editors and organizations that already exist in Librario. Depending on what they sent, this used to create a duplicate or drop the entry. The API now links the existing entry of your account to the publication. If a request lists the same person several times as author or editor, the publication shows them once. If a script names a person who does not exist in your account, the API reports an error instead of quietly saving something else. This applies to API v2 and v3.

Your scripts name the existing entry by its ID or its name. The details are in the API reference linked from the API documentation.

Disconnecting an AI assistant from your account used to also risk revoking the REST API access of the same application, for example the credential your own scripts ran on. The REST API row then vanished from the page with no explanation, and it worked the other way round too. The two are now independent: revoking one leaves the other untouched.

The Integrations page therefore lists every application that accesses Librario on your behalf under “Connected applications”, each labelled “AI assistant” or “REST API” and independently revocable. The confirmation prompt now names what you are about to break: your scripts and integrations for a REST API credential, running chat sessions for an AI assistant.

All connections also show when they were last used, or “Never used” if no request ever arrived, and one that has been used but not in the last 30 days carries an “Unused for over 30 days” note. That makes credentials nobody needs any more easy to spot before you revoke them. How to review and end connections is described in the AI assistants documentation.

Connecting an AI assistant to Librario used to put a list of tool names in front of you, so deciding what to allow meant knowing what download_file or get_publication actually do. The consent screen now shows one permission per line in plain words: “Search and view publications” and “Download and read attached files”. The second one combines what used to be two separate checkboxes, one for downloading and one for reading file content. What each checkbox does is spelled out in the AI assistants documentation.

Already-connected AI assistants: if you had allowed a connection only one of the two, downloading or reading file content, it does not get the combined permission automatically, since that would have allowed it more than you had consented to. It therefore loses the file capability it did have until you reconnect it under Account → Integrations and grant the permission again. Connections you had allowed both, or neither, keep exactly the access you had given them.

Scripts using the REST API also need permissions of their own for attached files: one to download them, one to upload, change or delete them. Permission to read or change library data is not enough. Existing REST API access received the matching file permission automatically; new scripts ask for it explicitly. Which permission allows what is listed in the API documentation.

When you connect an AI assistant to Librario, download_file hands it a signed link to the PDF — but for at least one major AI tool, that isn’t enough. Claude.ai’s own safeguard against autonomously fetching links that came out of a tool call means it won’t follow that URL itself, so the file content never reached the model. The assistant could show you the link, but not read the file, summarize it, or answer a question about its contents without you opening it yourself first.

Librario’s MCP server now also exposes publication details and file content as MCP resources — content delivered over the same already-authenticated connection your assistant has, not a link it has to fetch on its own. Ask your assistant to look up a publication or read one of its attached files, and on any AI tool that supports MCP resources, the content comes back directly in the conversation. get_publication and download_file keep working exactly as before for tools that prefer a link.

This is a new permission, not an automatic upgrade to what your assistant could already do: reading file content directly needs its own tick on the consent screen, in addition to the existing permission to download files. Untick just the direct reading and download links keep working exactly as before. Decline file downloads altogether and both go away — the links and the in-conversation reading.

For accounts with advanced AI controls, administrators can also switch off AI file-content access account-wide, independent of what any individual user has consented to. With it disabled, download_file no longer returns a link — it points assistants to the publication’s page for the person to open themselves instead — and the file-content resource is refused outright.

In Librario you can import books by their ISBN. The most tedious step used to be typing the 13 digits from the back of the book, often with a transposed digit. From now on you simply scan the barcode with your camera. Open the scanner with the Scan barcode button on the import page, hold the back of the book in front of the camera and confirm the ISBN it recognises. The import then carries on as usual, including duplicate detection.

For an overview of all ways to import, see Importing publications.

Since June, members place holds themselves when every copy of a publication is on loan. We’re now building on that. Until now, nobody heard automatically when a copy with a hold came back: the library team checked by hand who was next, and the person waiting only got a message once the team lent them the copy.

When a copy of a publication with a hold is returned, Librario now notifies the library team automatically. The next person in the queue also gets a direct notification, by email and in the app, and can borrow the copy right away. Only when self-service borrowing is disabled does Librario skip this direct notification: the library team then handles the loan as usual, and Librario notifies the waiting person once the team records the loan for them. Whoever returns a copy always sees in the return dialog who is next in line.

Librarians now see the queue right on the publication: the new Holds tab next to the copies shows how many requests are open and who is waiting. After returning a copy with a hold, the team lands in that tab automatically and lends the copy to the next person in one click.

Members see their position in the queue both on the publication page and under My loan requests. In libraries with self-service borrowing, the hold dialog already announces the notification when the hold is placed. More in What happens when you return a copy?.

API scripts used to need your Librario password. For API v3, you give each script and integration its own access instead. It reaches only as far as you allow, read-only for example, and you revoke it at any time without changing your password. API v3 no longer accepts your password. API v2 keeps password sign-in until it shuts down at the end of 2026. How to set up access is described in the API documentation.

When every hard copy of a title is on loan, users used to have to ask or wait. Some libraries kept the queue in an Excel sheet. That means manual effort, it is error-prone, and it is opaque for users.

Now the publication page offers a Place hold action: users add themselves to the queue as soon as no hard copy is free. Librario collects the holds in the title-specific queue, which librarians and admins see under “Tasks → Loan requests”. No external list needed. In “My Loan Requests”, users keep track of their holds and can withdraw them at any time.

Requesting and holding now share one term: the loan request. This feature used to be called “internal mail order”; the new name fits better, because a loan request now covers more than delivery via internal mail.

Some teams coordinate their work through a shared mailbox (e.g. library@…) rather than personal addresses. Until now Librario always assigned a newly created task to a single librarian. In a team like that, the notification should reach everyone, so that someone picks it up.

For this there is now a strategy under Automatic assignment in the account settings: “Notify shared email address(es)”. Librario then sends new tasks to one or more shared addresses without assigning them to a single person. If a task sits unattended for too long, a reminder goes to the same address list.

The strategy works for every auto-assignment category, not just user requests. A typical example is the loan request: when a user requests a hard copy, the request lands straight in the team’s shared mailbox.

A collection bundles the literature on a subject area or project. Choosing one already narrows down which publications are worth looking at. Even within that selection it is often unclear which title answers your actual question. Until now you could not find out: there was no search inside a collection.

The detail page of a collection now has a search field for exactly that. Enter a search term and Librario searches only the publications in that collection. It is the full library search, limited to the collection.

From the search field you land on the publications list, where the collection appears as a filter. Refine the result with all the usual filters and facets, or remove the collection filter with one click and search all your holdings again.

Documentation: managing collections

You can now upload your own cover image for any publication — right in the edit form, with a preview shown as soon as you pick a file. This gives you full control over which image appears on the publication card and the detail page.

Until now, Librario resolved the cover automatically: from the first page of an attached file, or by ISBN from online sources. That still works. If you do not upload your own image, Librario keeps looking up a matching cover for you and stores it.

What is new is the choice: when the automatically resolved cover does not match the edition at hand, simply upload your own image — it then takes precedence.

Users no longer research only in the web interface. They also have AI assistants such as Claude or ChatGPT help with literature research.

Until now that meant copying search results by hand, downloading PDFs and dragging them into the chat window, or letting the AI answer without access to your own holdings. Maintaining API keys or duplicating library data is out of the question for most librarians.

Librario now offers an MCP server (Model Context Protocol). With a one-off OAuth sign-in you connect your Librario account to your AI tool. From then on, the tool can search publications, fetch details and download files on your behalf, with the same permissions you have in the web interface.

On the consent screen you decide, permission by permission, what the assistant may do. If you want to allow only search and detail lookup, you switch off file downloads with one click. Your active connections are listed in the user menu under Integrations, and you can revoke individual connections there at any time.

You’ll find instructions for Claude.ai, Claude Desktop and ChatGPT, example requests and notes on security and data protection on the help page for the MCP integration.

In addition to the familiar BibTeX and EndNote exports, you can now export publications, collections and favourites as a CSV file. It suits Excel and data analysis tools.

The export contains all hits of the current search or collection, not just the first page. The column headings appear in your language (German or English) and match the field names in the form.

Librario now also indexes plain-text attachments: .txt, .md, .csv, .log and more.

Full-text search therefore finds matches not only in PDFs but also in notes, source data, log excerpts and accompanying texts attached to a publication.

Many publications are not dated to the day: for a monograph you often know only the year, for a journal the year and month. The date field, however, used to insist on day, month and year.

Librarians had to put an arbitrary “01.01.” (1 January) in front of the year, which skewed filter, sorting and export results and showed up in BibTeX and EndNote as an exact date.

Publication dates are now stored in the EDTF format, so 2024, 2024-03 and 2024-03-15 are all valid. Filters, the facets in the sidebar, the BibTeX, EndNote and Atom exports and the API all work with the precision that was actually entered.

EDTF stands for Extended Date/Time Format, an open standard from the Library of Congress for structured dates of varying precision.

We’ve extended our support for the parent publication best practice.

The publications list now has a Parent publication filter that shows all volumes of a multi-volume work, a series or conference proceedings at a glance. From the parent publication, a Show all related link leads straight to the filtered list, and from any volume you reach its siblings through the parent publication. You no longer have to gather them by hand.

You can now see the contract between your organisation and Librario directly in Librario, without searching your inbox or mail archive.

Librario holds:

• the current General Terms and Conditions
• the Data Processing Agreement (DPA) for the GDPR
• changes to these documents, so the account administrators can trace what was signed and when

You’ll find the documents under Account → Manage contracts.

The bibliographic field known as the “uniform title” is now called Original title in Librario. Inline help in the form explains, with an example, when and how to use the field.

On the detail page, the original title is clickable: one click filters the publications list to every edition of the same original title. That brings translations, editions and reprints of the same work together in one step.

On a publication’s detail page, the new Related collections card shows every collection that contains the publication.

Each collection is linked, and one click opens it directly. If the publication is in no collection, the card is hidden automatically.

In everyday library work, users email the library (“Could you get hold of this proceedings volume?”, “This DOI is missing”, “Could I borrow the copy?”). Librarians keep to-do lists, Excel sheets and sticky notes on the side.

As holdings grow, the inbox becomes critical infrastructure. Requests get lost, the status lives in one person’s head, and data quality problems (duplicates, missing identifiers) are visible but nobody prioritises them.

Librario now offers built-in task management with a Kanban board and a list view. Tasks are created automatically from data quality notices, loan requests, error reports and duplicate reviews, or you can add them by hand. Every task carries a due date and a priority, and can be commented on and assigned to colleagues. In-app notifications and an email summary keep people informed about assigned tasks.

Tasks replace the earlier email notifications: where users used to email the library an error report or a loan request, a traceable task now appears directly on the Kanban board, visible to the whole team with a status and due date, instead of disappearing into an inbox.

Basic task management is enabled for all accounts. Advanced task management is an optional add-on and includes your own recurring tasks, quality-review tasks after every import, automatic assignment by area of responsibility, due-date reminders and the quality assurance card on publication detail pages.

More detail in the documentation:

• The Kanban board: columns, card layout, filters, list view
• Task types: quality review, loan request, error report, general tasks
• Recurring tasks: define routines such as stocktake, standards maintenance or cataloguing new journal issues once

A publication could previously carry only one ISBN, one DOI, one PMID and so on.

In practice many publications have several, such as a hardcover and a softcover ISBN, or a print and an online DOI. Librarians had to settle on one variant or squeeze the second ID into a notes field, where it was lost to search and export.

Every publication can now carry any number of identifiers as an ordered list: ISBNs, DOIs, PMIDs, arXiv IDs, ISSNs and custom identifier types. Set the order by drag and drop. Search, CSV export and the API take every value into account; BibTeX and EndNote contain the first ISBN, DOI, ISSN and PMID.

API v3 is available as a preview release — the next generation of the Librario API. v3 clears out the legacy of v2 and brings, among other things:

• a typed identifiers list (instead of the flat doi/isbn/pmid/arxiv_id/issn fields)
• dedicated endpoints for nested resources (authors, editors, categories, identifiers) with replace semantics
• dedicated DELETE /…/:id endpoints instead of _destroy markers
• a unified PATCH /users/:id (instead of three different user update endpoints)
• the deprecated import_state field on publications is gone

Migration note: v3 is marked as a preview — the interface can still change before the stable release. For production use we recommend planning the migration now and carrying it out once v3 is stable.

API v2 shuts down at the end of 2026. Until then v2 stays available unchanged; individual fields (import_state, the flat identifier parameters, two of the user endpoints) already carry X-API-Deprecation headers. If you run v2 in production, plan the move to v3 over the coming months.

The complete v3 interface is described in the REST API documentation.

You can now record publications by their arXiv ID, both in the entry form and in the import assistant. Librario accepts the modern format (arXiv:2301.12345) and the older subject-based schemes (for example math.AG/0309136), and normalises the spelling automatically.

When you import by arXiv ID, Librario also fetches the metadata straight from arxiv.org: title, authors, abstract, publication date, DOI and journal details are filled in automatically, just as with an import by DOI.

When you record a publication or use the import assistant, Librario suggests common standards after two characters, for example DIN EN ISO 9001, EN 1991-1-4 or DIN 18960.

That saves typing and avoids spelling variants (“DIN-EN” instead of “DIN EN”). The same standard then looks the same in all ten records, which improves search and duplicate detection.

By default, the upload limit is 200 MB per file. That is enough for the vast majority of PDFs. We recommend keeping PDFs small so they stay quick to open for users, and optimising them for full-text search; see the best practice Optimising files for full-text search.

In special cases (for example high-resolution proceedings volumes or scanned holdings catalogues) some customers could not do this optimisation. Splitting files was a nuisance for users and counterproductive for research.

On request, support raises the upload limit for the whole account to up to 5 GB per file. The Import PDF files page then shows “unlimited” instead of “200 MB”.

Until now, the invoice went to exactly one email address, usually that of the person who created the account.

In many organisations that is not enough: invoices also have to reach a central invoice inbox (for example invoices@…) so that accounting can book the document, and the budget holder can assign the costs to their cost centre. Forwarding by hand was routine.

Every new invoice now goes automatically to all the addresses you have saved, so the invoice inbox and the budget holder are informed at the same time.

Under Our account > Subscription > Edit billing address, the Email addresses field takes up to five addresses (separated by a comma or a semicolon).

When you import several PDF files at once, you can now set the category and collection in the Apply to all publications block.

Imported records used to need moving into the right category and adding to a collection one by one, which quickly became a full day’s work for larger imports. Both fields now apply to every record in an import.

The import has been fundamentally reworked.

The new version gives you an overview of running and finished imports. Users now see straight away when an import has failed and why. You can retry or cancel a failed import with one click.

For a cancelled import, you can simply carry on editing the publications already created. Librario also detects duplicates automatically by identifier (ISBN, DOI, PMID, arXiv ID or standard number) and helps you add copies after the import.

The new import was first in beta and open to participants in the Early Adopter programme.

Documentation: Managing publications

You can now export all users as a CSV file and compare them with your staff directory, for example by employee number. That shows you the profiles of employees who have left, so you can clean them up.

The file contains ID, email address, first and last name, employee number, roles, phone number and a direct link to the user profile. You can process the data further in any common spreadsheet application.

You can find the export under Our account → Users: open Export at the top right and choose Export as CSV.

Documentation: User management

You can export single publications, search results and collections in BibTeX and EndNote format. That lets you use the data from Librario in reference managers such as EndNote, Citavi and BibTeX.

Until now, exporting search results and collections was limited to the first page of results.

That limit is gone. You now export all search results and all publications of a collection (every page) in one go as an EndNote or BibTeX file.

That makes the integration with common reference managers even closer.

Note: If you are thinking of exporting the whole library for specific analyses, we recommend our modern REST API. It integrates better with IT applications and scripts.

The improved export was first rolled out through the Early Adopter programme and is now available to all accounts.

You can now link publications hierarchically with the Parent publication field.

This groups the volumes of a multi-volume publication, and users can find all parts of a book more easily.

Example:

Take the multi-volume book “Beton-Kalender 2024” with parts 1 and 2. You would create three publications in Librario:

• “Beton-Kalender 2024”
• “Beton-Kalender 2024, Teil 1”, which has “Beton-Kalender 2024” as its parent publication
• “Beton-Kalender 2024, Teil 2”, which also has “Beton-Kalender 2024” as its parent publication

Documentation: Bibliographic fields

Best practices: Grouping publications

Publications now have an Original title field.

The original title (in library terms: uniform title) is the title that ties together several editions of a publication.

One use is grouping different editions of the same publication. The original title is the basis for pointing to newer editions later.

Another use is grouping the different language editions of the same work. That makes it easier for users to find other language editions.

Documentation: Bibliographic fields

Best practices: Grouping publications

After more than ten years, Librario has a new user interface. Several areas were rebuilt and simplified.

Users can create collections of publications themselves and share them with other users.

Collections serve different purposes:

• An engineering firm uses them to model its “Handapparate”, the working collections of literature its teams keep to hand. Specialists maintain curated collections of the most important literature on a subject. That helps new colleagues get up to speed on a topic quickly, or communicates the company’s “state of the art” clearly.
• A medical technology company uses collections to share the results of partly outsourced literature research with its scientists.

Documentation: Managing collections

Librario is now a Progressive Web App.

With Chrome on Windows, Mac and Android you can install Librario as an “app”. Librario also keeps working with an unstable or missing internet connection: pages you have opened before load from the cache.

Documentation: Progressive Web App (PWA)

Until now we used Google Analytics to see how Librario is used day to day, so that we could improve the product.

We have decided to stop using Google Analytics.

This benefits data protection and also speeds up the web application for all users.

Uploaded files could previously be 150 MB at most. We have raised the limit to 200 MB.

We advise against using Librario for files this large. Nobody opens one casually just to look something up, and the download takes too long.

It is worth reducing the size of such files. Modern PDF editors can help.

There are now more reports that help librarians raise the quality of the library.

Librario can now identify possible duplicates by DOI or ISBN.

For each possible duplicate, librarians decide whether it really is a duplicate publication (then one of the two can be deleted) or whether there is another reason for the shared ISBN or DOI. One possible reason: a multi-volume work in which each volume is stored as its own publication. Multi-volume works often share one ISBN.

*[DOI]: Digital Object Identifier *[ISBN]: International Standard Book Number

When someone searches for a contributor’s surname, Librario now suggests filtering by that person instead.

You can now share a publication with your colleagues from its detail page.

Besides categories, you can now add keywords to publications in Librario.

Keywords make a publication easier to find under particular terms, such as synonyms.

Categories build a hierarchy of publications that you can filter by when you search.

Documentation: Bibliographic fields

Librarians now have a dashboard. It suggests how to improve the quality of your library’s data.

Why does the quality of your holdings matter?

Search in Librario can only be as good as your data allows. That is why it pays to fix typical quality problems.

The language of a publication, for example, lets Librario apply language-specific optimisations to full-text search. The year is useful too: many users treat it as an important relevance criterion when they search. Some users show only the most recent publications. Publications without a year would then be left out.

There is now a dedicated page for managing the location information of your hard copies.

Librarians can rename a location or, if needed, merge two locations, for example different spellings of the same place.

Documentation: Managing hard copies

Most objects in Librario now have a REST API, so you can read, create and edit data.

It is aimed at IT departments and teams with their own development who retrieve or create Librario content automatically.

Documentation: REST API

There are two new publication types:

• Patent (patent): for patent specifications
• Periodical (periodical): for individual issues of a journal or magazine

The publication type conference is now called Proceedings (proceedings), because that is the more common term.

Documentation: Publication types

By default, Librario sorts search results by relevance. Relevance comes from a score that Librario calculates internally for your search term.

Some users prefer to see the newest publications first. There is now an option for that: Chronological order.

We have completely reworked the search.

What is new:

• Better results through better sorting by relevance
• You can now search by ISBN and DOI
• New filters: filter for publications where a particular field is empty
• Optional full-text search: you decide whether Librario searches only titles and metadata or the full text as well
• Search your personal favourites separately
• Besides German and English, the search index now supports French, Spanish and Italian

Documentation: Searching for publications

Not every specialist article matters equally to all employees: it depends on their training focus and specialisation in day-to-day work.

That is why everyone can now keep their own list of favourites.

Each publication also shows who else has it among their favourites. A user’s profile lists their favourites next to the books they have borrowed. So colleagues quickly discover new and relevant literature.

Documentation: Managing favourites

We have spent the last few weeks tuning Librario under the hood: it responds faster, and several minor bugs are fixed. This rests on more accurate measurement of response times.

When you edit a publication, the cover image is now shown larger. You see at a glance which publication you are working on.

There is also a new button to download the linked files directly. Search responds faster again.