The Abstractinator MCP server
Commit Nov 15, 2026Enables AI assistants to search immunology and virology literature, clinical trials, and research grants across multiple databases simultaneously.
Also known as ckangas2/abstractinator, abstractinator
- Stars
- 0
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- 2.1.0
- Last commit
- Nov 15, 2026
- License
- —
What is The Abstractinator MCP?
The Abstractinator enables AI assistants to search immunology and virology literature, clinical trials, and research grants across multiple databases simultaneously.
Summary generated from the project's README
About The Abstractinator MCP server
Researchers in immunology and virology can perform comprehensive literature sweeps through a single interface. This tool queries multiple databases simultaneously, including Europe PMC, OpenAlex, ClinicalTrials.gov, bioRxiv, medRxiv, NIH RePORTER, and NSF Awards. It automatically removes duplicate records by DOI and uses a scoring matrix to tag every result with the most relevant immune cell type, virus, or bacterium.
Developers can connect AI assistants like Claude to the service using the remote MCP server via a custom connector. The server provides specific tools such as search_literature for retrieving records, get_fulltext for accessing open-access articles, and program for looking up specific therapy programs. While the AI connector uses open-access sources by default, users can provide their own API keys for CORE, USPTO, or Scopus to expand coverage.
This system is designed for speed, with standard searches completing in 5 to 10 seconds. It also features a deep search mode that increases coverage by pulling up to 250 results per source. By integrating directly with AI clients, it allows for automated discovery of clinical trials, research grants, and scientific papers.
What you can do with it
- 1Simultaneous multi-database literature sweepsSearch across several biomedical and grant databases at once to find immunology research.
- 2Automated biological entity taggingIdentify the specific viruses or immune cells discussed in a research paper automatically.
- 3Therapy program trackingRetrieve all linked clinical trials, papers, and NIH grants associated with a specific therapy code.
- 4Clinical trial change monitoringTrack updates to clinical trials such as new results or changes in trial phase.
Written from the project's README. Every use case is checked against a quote from it.
How to install The Abstractinator MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall The Abstractinator MCP in Claude Code
$ claude mcp add --transport http abstractinator https://mcp.abstractinator.me/mcpInstall The Abstractinator MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"abstractinator": {
"url": "https://mcp.abstractinator.me/mcp"
}
}
}Install The Abstractinator MCP in VS Code
.vscode/mcp.json
{
"servers": {
"abstractinator": {
"type": "http",
"url": "https://mcp.abstractinator.me/mcp"
}
}
}Install The Abstractinator MCP in VS Code CLI
$ code --add-mcp '{"name":"abstractinator","type":"http","url":"https://mcp.abstractinator.me/mcp"}'Install The Abstractinator MCP in Codex
$ codex mcp add abstractinator --url https://mcp.abstractinator.me/mcpInstall The Abstractinator MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.abstractinator] url = "https://mcp.abstractinator.me/mcp"
Install The Abstractinator MCP in Gemini CLI
$ gemini mcp add -s user --transport http abstractinator https://mcp.abstractinator.me/mcpInstall The Abstractinator MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"abstractinator": {
"url": "https://mcp.abstractinator.me/mcp",
"type": "http"
}
}
}Configuration
ABSTRACTINATOR_LIVE_TESTSDATA_WEBHOOK_URLENABLE_SCOPUSINTEGRATE_NIH_NSFOPENALEX_API_KEYUSER_EMAIL
| Name | Description | |
|---|---|---|
| ABSTRACTINATOR_LIVE_TESTS | — | |
| DATA_WEBHOOK_URL | — | |
| ENABLE_SCOPUS | — | |
| INTEGRATE_NIH_NSF | — | |
| OPENALEX_API_KEY | — | |
| USER_EMAIL | — |
Automated checks
Ran Nov 15, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableckangas2/abstractinator is public and reachablegithub.com/ckangas2/abstractinator ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)No licenseFinding
- Commit or release in the last yearLast commit or release 2026-11-12github.com/ckangas2/abstractinator ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeinitialize answered with protocol 2025-06-18 in 414 msPass
README
View on GitHubThe Abstractinator turns a literature sweep into one search. It queries Europe PMC, OpenAlex, ClinicalTrials.gov, bioRxiv/medRxiv, NIH RePORTER and NSF Awards in parallel, removes duplicates across sources, and tags every record with its most prominent immune cell type and virus using a weighted title/abstract scoring matrix.
A new search takes about 5–10 seconds; repeat searches within 7 days are instant.
✨ Features
| Searchinator | Queries every source simultaneously; the slowest source sets the pace, not the sum of all of them. |
| Deep search | Optional mode that pulls up to 250 results per source instead of 50 (~4× the coverage, ~15–30 s). |
| Deduplicatinator | Merges records across sources by DOI (including preprint → published DOI links), then by title, preferring the best-curated source. |
| Tagging | Labels each record with the immune cell type, virus and bacterium it is most strongly about, from curated alias lists. |
| Plotinator pools | Switch the plot's pathogen axis between viruses and bacteria (or any pool you add) without re-running the search. |
| Plotinator | Interactive charts of viral–immune trends over time. |
| Readinator | Build a reading list and export it to Zotero or EndNote. |
| Agentinator | An MCP server so AI assistants (Claude and others) can search The Abstractinator directly. |
📚 Data sources
| Source | Access | Coverage |
|---|---|---|
| Europe PMC | Open | Biomedical literature (incl. PubMed) |
| OpenAlex | Open | Global scholarly graph |
| ClinicalTrials.gov | Open | Human clinical study registrations |
| bioRxiv / medRxiv | Open (local copy) | Preprints, searched from a local Parquet store refreshed daily |
| NIH RePORTER | Open | US federal grants |
| NSF Awards | Open | Fundamental research grants |
| CORE | Your API key | Open-access full-text aggregation |
| USPTO (PatentsView) | Your API key | Patents |
| Scopus / Embase | Your API key, off by default | Abstract retrieval only works from a subscribing institution's network, so it's disabled unless ENABLE_SCOPUS=1 |
Keyed sources are bring-your-own-key: enter keys in Settings (☰). They're held only in your browser session, never logged or written to disk, and never shared with other visitors.
🤖 Use it with AI
The Abstractinator is available as a remote MCP server, the open standard AI assistants use to connect to tools.
https://mcp.abstractinator.me/mcp
In Claude: Settings → Connectors → Add custom connector → paste the URL above, and choose No sign-in for authentication. Then just ask, e.g. "Search the Abstractinator for oncolytic virus trials in melanoma." Any MCP client that supports remote (streamable HTTP) servers works too.
The server exposes four tools:
search_literature(query, max_results=25, abstract_chars=1500, deep=False, offset=0, sources="")
returns total_found, available, has_more, source_counts, unavailable_sources, and a
source-balanced list of records with title, abstract, authors, publication_date, doi,
pmid, url, source, primary_cell, primary_virus, primary_bacteria, primary_modality,
is_bioinformatics, is_retracted, for open-access articles pmcid and is_open_access, and for
clinical trials trial_status, trial_start, trial_end, phase, interventions, intervention_other_names, conditions, lead_sponsor and trial_references. Every record has every key.
get_fulltext(pmcid, doi, max_chars=20000) returns the open-access full text of one article.
program(query) looks up one therapy program in the program registry by its stable AXP ID or
any name or code, and returns its aliases, linked ClinicalTrials.gov trials (each with how it was
linked, a confidence, and auto or unreviewed candidate status), the papers (PubMed, clinical first)
and NIH grants (RePORTER) that name it, linked the same way, and recent changes.
changes(since, program, modality, significance="info", limit=100) returns change events
(new trials, status and phase changes, results posted...) from weekly ClinicalTrials.gov snapshots
of the registry's trials.
Tag agreement with NCBI's PubTator3 is measured in benchmark/pubtator.md, on a held-out set of searches that name no organism as well as on the searches used to tune the alias lists. The figures are agreement with another model, not accuracy.
The AI connector uses open-access sources only (no keys needed or spent) and is rate-limited to keep the service available for everyone.
🏗️ How it works
Browser ──► Cloudflare ──► abstractinator.me ──► Shiny Server :3838 (app.R) ─────────┐
├─► search engine ──► shared cache
AI ──► Cloudflare ──► mcp.abstractinator.me ──► MCP server :8200 (agent/mcp_server.py)│ (orchestrate_extraction.R)
└──► REST API :8100 (agent/api.R) ┘
- One search engine, two front doors.
orchestrate_extraction.Rchecks the cache, starts every source as a background job (future), then combines, deduplicates and tags the results. The website calls it asynchronously through Shiny'sExtendedTask, so one visitor's search never freezes the page for anyone else. The REST API calls the same code synchronously. - Shared cache. Results are cached for 7 days in
.cache/s3_mimic/(one.rdsper search, keyed by term, depth and which keyed sources were used). A search by a person makes the same search instant for an AI, and vice versa. - Local preprints. bioRxiv/medRxiv are searched from
biorxiv_local_db/(one Parquet file per year) via Arrow, instead of hitting their API on every search. - Hit detection. Titles and abstracts are matched against immune-cell, virus and bacteria
alias lists (
aliases.R); title hits weigh 20, abstract hits 3, and the top scorer becomesprimary_cell/primary_virus/primary_bacteria. A fourth axis,primary_modality, tags the therapy (oncolytic virus, CAR-T, checkpoint inhibitor, vaccine, chemotherapy...), counting a trial's registered interventions like its title. Aliases are matched as whole words, short acronyms case-sensitively, and context rules drop phrases like "ad hoc" or "CMV promoter". - Therapinator (in development, hidden on the website:
SHOW_THERAPINATORinapp.R). "What's been tried": a table of therapy x stage (grant, preprint, paper, trial phase) for the current results, and the trials themselves with their interventions, phase and status, most advanced first. The therapy tags themselves (primary_modality) are live in the API. - Swappable pools. The Plotinator's pathogen axis is not hard-wired to viruses: a pool switch
selects which tagged category it plots, and the same search can be viewed through either lens.
Adding a pool (fungi, parasites, drug classes, a lab's own vocabulary) means defining an alias
list in
aliases.R, tagging it withtag_with_alias_list()inorchestrate_extraction.R, and adding one entry toPATHOGEN_POOLSinapp.R— contributions welcome. - Logging. Website searches write a small JSON record (term, duration, result count, cache
hit, and whether keys were present, never the keys) to
.cache/s3_mimic/logs/. Feedback submitted through the app is saved to.cache/s3_mimic/feedback/and, ifDATA_WEBHOOK_URLis set, forwarded to Discord. - Display safety. Text from external databases is HTML-escaped before display, with only
basic formatting tags (italics, bold, sub/superscript, line breaks) allowed through, and links
restricted to plain
http(s)URLs.
🖥️ Self-hosting
Requirements
- Linux (tested on Ubuntu), R ≥ 4.1, Python ≥ 3.10 (for the MCP server)
- System libraries:
sudo apt install -y r-base r-base-dev libcurl4-openssl-dev libssl-dev libxml2-dev \ libsqlite3-dev libfontconfig1-dev libharfbuzz-dev libfribidi-dev libfreetype6-dev \ libpng-dev libtiff5-dev libjpeg-dev libglpk-dev libgmp3-dev libsodium-dev cmake gfortran - R packages (install into the system library so Shiny Server can see them):
sudo /usr/bin/Rscript -e 'install.packages(c("shiny","bslib","shinyjs","shinythemes","shinyWidgets", "tidyverse","DT","plotly","htmltools","htmlwidgets","htmlTable","jsonlite","digest","future", "promises","parallelly","arrow","httr","httr2","RSQLite","DBI","uuid","lubridate","ggbeeswarm", "plumber"), repos = "https://cloud.r-project.org", Ncpus = 8)'
Using conda? Shiny Server and the services use the system R at
/usr/bin/R. Install packages with/usr/bin/Rscript(orconda deactivatefirst) so they land in the right place.
1. Get the code and configure
git clone https://github.com/ckangas2/abstractinator.git
cd abstractinator
Optional .Renviron in the project root:
INTEGRATE_NIH_NSF=1 # include NIH & NSF grants (the app also sets this)
USER_EMAIL=you@example.org # sent to OpenAlex for its "polite pool"
OPENALEX_API_KEY= # free key from openalex.org: 10x the daily allowance, needed for deep search paging
ENABLE_SCOPUS=0 # set to 1 only on a subscribing institution's network
DATA_WEBHOOK_URL= # optional Discord webhook for updater notifications
No paid API keys are needed on the server: the website is strictly bring-your-own-key and
never uses paid-source keys from .Renviron, and the AI connector uses open-access sources
only. The one server key is the free OPENALEX_API_KEY, which raises OpenAlex's daily
allowance for everyone's searches; it is sent as a request header, never logged, and never
reaches the browser. Without it, a busy day can exhaust the keyless allowance (HTTP 429). (Keys in
.Renviron are only read by command-line scripts you run yourself.)
2. Build the preprint store (once, takes several hours)
tmux new -s biorxiv
Rscript write_parquet_bioRxiV.R # resumable: finished years are skipped on rerun
Detach with Ctrl+B, D. Searches pick up each year as soon as it's written.
3. Run locally
/usr/bin/Rscript -e 'shiny::runApp(port = 4000)' # website
/usr/bin/Rscript agent/run_api.R # REST API on 127.0.0.1:8100 (run from repo root)
4. Production setup (what abstractinator.me runs)
Shiny Server, serving the app at the root and only on localhost
(/etc/shiny-server/shiny-server.conf):
run_as shiny;
server {
listen 3838 127.0.0.1;
location / {
app_dir /srv/shiny-server/abstractinator;
log_dir /var/log/shiny-server;
app_idle_timeout 0;
}
}
API and MCP services (run as the shiny user):
sudo /usr/bin/python3 -m venv /opt/abstractinator-mcp
sudo /opt/abstractinator-mcp/bin/pip install -r agent/requirements.txt # pinned, MCP Python SDK 2.x
sudo cp agent/abstractinator-api.service 'agent/abstractinator-api@.service' agent/abstractinator-mcp.service /etc/systemd/system/
sudo systemctl daemon-reload
sudo systemctl enable --now abstractinator-api abstractinator-mcp
Daily preprint updates (after the initial build finishes):
sudo -u shiny mkdir -p logs
sudo cp deploy/abstractinator.cron /etc/cron.d/abstractinator # runs update_biorxiv_db.R at 4:15am
Public access via a Cloudflare Tunnel
(no open ports): route abstractinator.me → http://localhost:3838 and
mcp.abstractinator.me → http://localhost:8200. The REST API stays private.
Updating the live site
cd /srv/shiny-server/abstractinator
sudo -u shiny git pull
sudo systemctl restart shiny-server # app.R, www/, search scripts
sudo systemctl restart abstractinator-api # agent/api.R, search scripts (all three API processes)
sudo systemctl restart abstractinator-mcp # agent/mcp_server.py
Running tests
# Fast offline tests (these also run on GitHub with every push)
Rscript -e 'testthat::test_dir("tests/testthat")'
# Plus live end-to-end searches against the real databases (needs internet, ~1-2 min)
ABSTRACTINATOR_LIVE_TESTS=1 Rscript -e 'testthat::test_dir("tests/testthat")'
Usage logs
Every search, from the website or an AI agent, is logged without personal data. Searches per day, split by channel:
cd /srv/shiny-server/abstractinator
sudo -u shiny /usr/bin/Rscript -e 'source("R/log_utils_local.R"); print(summarize_search_logs())'
REST API reference (local only)
GET /search?q=<term>&limit=25&abstract_chars=1500&deep=false
limit: results returned (1–100), picked round-robin across sourcesabstract_chars: truncate abstracts;0omits themdeep:truefor up to 250 results per source- Optional headers
X-CORE-KeyandX-USPTO-Keyenable those sources for the request - Server-wide limit of 20 searches/minute (HTTP 429 when exceeded)
GET /program?id=<AXP ID or name>
- One program from the registry database (
/var/lib/abstractinator/registry/registry.sqliteon bigtop; set by the service unit, override withABSTRACTINATOR_REGISTRY_DB): its virus or vector (for virus-based programs), approvals, aliases, linked trials, recent changes; a merged ID resolves to its survivor (merged_from) papers(up to 200;pmid,title,journal,year,kind,link_method,link_confidence,link_status,nct_id; auto links first, then newest) andgrants(up to 100, newest first;core_project_num,title,org,pi,first_fy,last_fy,urland the link fields); details not yet fetched arenull. Layout and rules:tracker/REGISTRY_DB.md- An ambiguous name returns
"program": nullandcandidates; no match is a 404, and a registry not built yet a 503
GET /changes?since=YYYY-MM-DD&program=&modality=&significance=info&limit=100
- Trial change events seen on or after
since(default 30 days ago), newest first significance:majoronly,minor(major and minor) orinfo(all);limit1–500/programand/changesshare a limit of 120 requests/minute
GET /health returns {"status":"ok"}.
🗂️ Repository layout
| Path | What it is |
|---|---|
app.R |
The Shiny web app: setup, and the order its parts load in |
app/ui.R, app/dt_options.R |
The page layout and the results table's options |
app/server/*.R |
The app's server, one file per area (search, plot, reading list, ...) |
orchestrate_extraction.R |
The search engine: cache → parallel sources → dedup → hit detection |
R/engine.R |
The list of files that make up the engine, loaded by the app, the API and the tests |
*_extraction.R |
One extractor per source (EPMC, OpenAlex, CORE, ClinicalTrials.gov, bioRxiv, NIH, NSF, PatentsView, Scopus) |
deduplication.R, aliases.R |
Cross-source deduplication; immune cell & virus alias lists |
R/db_utils_local.R, R/log_utils_local.R |
Search cache and search logging (website and agent searches) |
R/display_utils.R |
HTML sanitizing, export text cleanup and stable result IDs for the app |
tests/ |
Offline unit tests and opt-in live search tests (testthat); run on GitHub Actions |
agent/ |
REST API (api.R, run_api.R), MCP server (mcp_server.py), systemd units |
deploy/ |
Cron schedule |
write_parquet_bioRxiV.R, update_biorxiv_db.R |
Build and refresh the local preprint store |
www/ |
Styles, scripts, icons and mascot art |
archive/ |
Retired earlier generations, kept for reference |
🛣️ Roadmap
The direction is from retrieval toward curation. The pipeline already applies consistent editorial rules — when two records are the same work, which version to keep, which vocabulary labels apply, and how to sample across sources — but it does not yet judge whether a study is sound or how relevant it is to the question asked. Each item below adds one more judgment.
- Relevance ranking (best match first, using hit scores and recency)
- Skip caching searches where a source failed
- Log AI-agent searches alongside website searches
- Smarter agent tools backed by a relational knowledge base (gap analysis, "what's under-studied")
⚖️ License & branding
The code is open source under the MIT License: you're welcome to use, modify and build on it, as long as the copyright notice comes along.
The program registry (registry/ and the pages built from it) is under
CC BY 4.0: reuse it for anything, with credit to "The Abstractinator
program registry".
The name and branding are not part of that license. "The Abstractinator", the -inator feature names (Searchinator, Deduplicatinator, Plotinator, Readinator, Agentinator) and the look of the site identify the official project at abstractinator.me. If you run your own copy or fork, please give it your own name and branding, and don't present it as the official Abstractinator. A "based on The Abstractinator" credit with a link back is appreciated.
Third-party icons in www/ belong to their creators and are used under the Flaticon license
(see CREDITS.md); they aren't covered by the MIT License.
🙏 Credits
Mascot and icons from Flaticon (narwhal by Smashicons, dinosaur by imaginationlol, frog prince by Magnific); line-art animations from the Noun Project. Details in CREDITS.md.
📜 Citation
If you use The Abstractinator in your research, please cite:
Kangas, C. (2026). The Abstractinator: Automated Boolean Search Tool for Rapid Acquisition, Classification, and Triage of Immuno-Networks And Therapeutic Oncolytic Research (Version 2.1.0) [Computer software]. https://abstractinator.me
@software{TheAbstractinator2026,
author = {Kangas, Chase},
title = {The Abstractinator: Automated Boolean Search Tool for Rapid Acquisition, Classification, and Triage of Immuno-Networks And Therapeutic Oncolytic Research},
year = {2026},
url = {https://abstractinator.me},
version = {2.1.0}
}
Reproducing the environment
renv.lock records the exact package versions the live deployment runs,
generated from that machine's library (R 4.6.1, 155 packages). To install the
same set:
install.packages("renv")
renv::restore(lockfile = "renv.lock", library = .libPaths()[1], prompt = FALSE)
The lockfile is a record, not an activated renv project: there is no
.Rprofile and no renv/ directory in this repository, both of which are
gitignored. The live server is updated with git pull, and an .Rprofile
arriving that way would switch the Shiny and plumber services onto a project
library that is not populated for the service user, taking the site down.
.github/workflows/tests.yml runs the offline suite on every push against
current package versions; .github/workflows/deps.yml restores the locked
versions weekly and runs the same suite, so drift shows up without slowing
every commit.
The Abstractinator MCP questions
How do I add The Abstractinator MCP to Claude Code?
Run claude mcp add --transport http abstractinator https://mcp.abstractinator.me/mcp in your terminal. Add --scope user to make it available in every project.
How do I add The Abstractinator MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
What databases does it search?
It queries Europe PMC, OpenAlex, ClinicalTrials.gov, bioRxiv/medRxiv, NIH RePORTER, and NSF Awards in parallel.
How do I use it with Claude?
In Claude, go to Settings → Connectors → Add custom connector, paste the provided URL, and choose 'No sign-in' for authentication.
Does the AI connector require an API key?
No, the AI connector uses open-access sources only and does not require any keys.