Skip to main content

Connector Examples

This page provides working examples for each connector available through the Alien platform. Each example shows the user prompt, which tools are invoked, and what output to expect. Use these as a starting point to understand what the connectors can do and how to get the most out of them.

OpenAIRE Connector​

The OpenAIRE connector provides access to the OpenAIRE Research Graph — 600M+ research products including publications, datasets, software, and other research outputs. It is available through the openscience plugin (see Alien Marketplace).

Example: Find Recent Papers on a Topic​

User prompt: "Find recent papers on CRISPR gene therapy published in the last two years and summarize the top results."

What happens:

  1. search_research_products is called with keywords CRISPR gene therapy and a date filter for the last two years.
  2. The top results are returned with titles, authors, DOIs, publication dates, and bibliometric indicators.
  3. get_research_product_details is called for the top 3-5 results to retrieve full metadata including abstracts.

Expected output: A summary of the most relevant recent publications on CRISPR gene therapy, including paper titles, authors, publication venues, and a brief description of each paper's contribution based on its abstract and metadata.

Example: Get Citation Network for a DOI​

User prompt: "Build a citation network for DOI 10.1038/s41586-020-2649-2 and identify the most influential citing papers."

What happens:

  1. get_research_product_details is called with the provided DOI to retrieve the seed paper's metadata.
  2. get_citation_network is called with the DOI as the seed to retrieve both citing and cited papers.
  3. find_by_influence_class is used to filter the citing papers by bibliometric class (C1-C3) to identify the most influential ones.

Expected output: A structured citation network showing the seed paper, its key references, and the papers that cite it. The most influential citing papers (by OpenAIRE influence class) are highlighted with their bibliometric indicators, helping the user understand the paper's impact and how the field has evolved since its publication.

Example: Compare Research Output Between Two Organizations​

User prompt: "Compare the research output of CERN and Fermilab in high-energy physics over the last five years."

What happens:

  1. search_research_products is called with keywords high-energy physics filtered to the CERN organization, with a date range of the last five years.
  2. search_research_products is called again with the same keywords filtered to Fermilab.
  3. analyze_research_trends is called for each organization to retrieve publication volume over time.
  4. find_by_influence_class is optionally used to compare the proportion of high-impact publications from each organization.

Expected output: A comparative summary showing publication counts, high-impact papers (by bibliometric class), and publication trends over time for both organizations. The user sees which organization is more active in the field, which has produced more highly-cited work, and how their outputs have trended over the five-year period.

Example: Discover Research Datasets on a Topic​

User prompt: "Find open research datasets related to climate change modeling that I can use in my analysis."

What happens:

  1. search_datasets is called with keywords climate change modeling.
  2. The results include dataset titles, descriptions, hosting data sources, access rights, and related publications.
  3. get_research_product_details is called for the most promising datasets to retrieve full metadata and linked publications.

Expected output: A list of openly accessible research datasets related to climate change modeling, including where each dataset is hosted, its access rights, a description of its contents, and links to the publications that created or used the dataset.

bioRxiv / medRxiv Connector​

The bioRxiv and medRxiv connectors provide access to preprint data clusters for life sciences and medical research. They use the same datacluster_* tool interface as your own data clusters. Available through the openscience plugin (see Alien Marketplace).

Example: Search for Preprints on a Topic​

User prompt: "Search for preprints about mRNA vaccine delivery mechanisms published on bioRxiv in the last year."

What happens:

  1. datacluster_list_datasets is called on the bioRxiv MCP server to confirm the available dataset.
  2. datacluster_keyword_search is called with the query mRNA vaccine delivery mechanisms and a date filter for the last year.
  3. The top results are returned with titles, authors, posting dates, and relevance scores.

Expected output: A ranked list of bioRxiv preprints matching the query, showing each preprint's title, authors, posting date, and DOI. The user can see the most relevant recent work on mRNA vaccine delivery without waiting for formal peer review publication.

Example: Read the Full Text of a Preprint​

User prompt: "Read the full text of the bioRxiv preprint with DOI 10.1101/2024.01.15.575700 and summarize its methods section."

What happens:

  1. datacluster_keyword_search is called with the DOI to locate the specific entry in the bioRxiv data cluster.
  2. datacluster_get_entry_content is called with the entry ID to retrieve the processed full text of the preprint (paginated if the content is large).
  3. If the full text spans multiple pages, additional datacluster_get_entry_content calls are made with incrementing page numbers until the methods section is fully retrieved.

Expected output: A summary of the preprint's methods section, extracted from the full processed text. The summary covers experimental design, techniques used, sample sizes, and analytical approaches — grounded entirely in the preprint's actual text rather than metadata alone.

Example: Find Figures from a Preprint​

User prompt: "Find figures from the paper on neural organoids by Lancaster et al. on bioRxiv and describe what they show."

What happens:

  1. datacluster_keyword_search is called with the query neural organoids Lancaster to locate the relevant preprint.
  2. datacluster_get_entry_documents is called with the entry ID to list all files attached to the entry, including extracted figures.
  3. datacluster_get_entry_file is called for each figure file (e.g., processed/figures/figure_1.png, processed/figures/figure_2.png) to retrieve the images.

Expected output: The figures from the preprint are retrieved and displayed as images. The AI agent describes each figure based on visual analysis — for example, microscopy images of organoid structures, growth curves, or immunostaining panels — providing the user with a visual understanding of the paper's key results without needing to download the full PDF.

Example: Semantic Search Across Preprints​

User prompt: "Find bioRxiv preprints that discuss similar methods to single-cell RNA sequencing of tumor microenvironments, even if they don't use that exact phrase."

What happens:

  1. datacluster_vector_search_chunks is called on the bioRxiv MCP server with the query single-cell RNA sequencing of tumor microenvironments. This performs semantic (meaning-based) search rather than keyword matching.
  2. The results return text chunks from preprints that are semantically similar to the query, along with similarity scores and source entry IDs.
  3. datacluster_get_entry_content is called for the top-scoring entries to retrieve additional context.

Expected output: A list of preprints that discuss conceptually related methods — such as single-nucleus RNA-seq, spatial transcriptomics of tumors, or immune cell profiling in cancer — even if they use different terminology. Each result includes the relevant text passage and a similarity score, helping the user discover related work they might miss with keyword search alone.

Combining Connectors​

The OpenAIRE and bioRxiv/medRxiv connectors can be used together in a single conversation. For example:

User prompt: "Find the most cited papers on CAR-T cell therapy using OpenAIRE, then check bioRxiv for any recent preprints from the same research groups."

What happens:

  1. search_research_products (OpenAIRE) is called with keywords CAR-T cell therapy, filtered by influence class C1-C3.
  2. Author names are extracted from the top results.
  3. datacluster_keyword_search (bioRxiv) is called with the author names to find recent preprints from those groups.
  4. Results from both sources are synthesized.

Expected output: A combined view showing the most impactful published papers on CAR-T cell therapy alongside the latest preprint activity from the same research groups — giving the user both the established literature and the cutting edge of the field in one answer.

Next Steps​