Summary

Cross-dataset search is the query the separate sources cannot answer.

PubMed, ClinicalTrials.gov and the FDA approvals list each have their own interface and their own idea of what a query means. Asking one question across all of them — what does the literature say about this drug, is it in trials, has it been approved — is the thing this makes cheap. The MAMMAL predictions are a different kind of output and should be read as model estimates with the confidence that implies, not as findings; the free tier's 100 requests an hour is enough for research and not enough for a batch job.

What it is

A remote MCP server over the Cure Cancer With AI public API. It puts five oncology datasets behind one connection — peer-reviewed papers from PubMed, curated news, FDA-approved oncology drugs, clinical trials from public registries, and a cross-dataset search — and adds three IBM MAMMAL foundation-model predictions on top.

What you get

A keyword search across every dataset at once, rather than four separate queries — `search_oncology`
Peer-reviewed oncology papers listed and fetched individually — `list_research`, `get_research_paper`
Curated cancer news and editorial blog posts, with excerpts on the list and full content on the single item — `list_news`, `list_blog_posts`, `get_blog_post`
FDA-approved oncology drugs, and clinical trials with eligibility and locations on the single-trial call — `list_fda_approvals`, `list_clinical_trials`, `get_clinical_trial`
Three MAMMAL model predictions: protein–protein interaction, drug–target binding affinity as pKd, and ClinTox clinical-trial toxicity, with a health check for the engine — `predict_ppi`, `predict_dti`, `predict_clintox`, `mammal_health`
List endpoints that page properly, with `limit` between 1 and 100 defaulting to 20, an `offset`, and filters on cancer type, search text and a date range

Requirements

A free API key — sign in at curecancerwithai.com/api-keys and create one; it is shown once and looks like `ccw_live_...`. Connect to `https://www.curecancerwithai.com/api/mcp` over Streamable HTTP with the key as a bearer token in the `Authorization` header. The free tier allows 100 requests per hour per key on a rolling window; watch the remaining-requests header and wait for `Retry-After` on a 429. Everything returned is informational and not medical advice, and the vendor asks that source URLs be cited when results are shown to a user.

Setup effort

One command plus a key — npx -y mcp-remote https://www.curecancerwithai.com/api/mcp, then supply credentials