PubMed / NCBI E-utilities
Provides programmatic access to NCBI databases including PubMed (35M+ biomedical citations), PMC full-text articles, gene sequences, and clinical trial records via the E-utilities REST API.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Read-only public bibliographic data; API key is a simple query parameter (not a header), which carries minor leakage risk in server logs. No sensitive data is accessible.
⚡ Reliability
Best When
An agent needs structured biomedical literature search with MeSH taxonomy, PMID-stable identifiers, and no authentication overhead.
Avoid When
You need full-text content from paywalled journals or real-time data refresh faster than NCBI's indexing cycle (typically 1–2 days).
Use Cases
- • Search PubMed for peer-reviewed papers on a drug interaction or disease and return structured citation metadata
- • Fetch full abstract and MeSH term annotations for a specific PMID to feed into a summarization pipeline
- • Retrieve linked records across NCBI databases (e.g., gene → protein → literature) for multi-hop research queries
- • Download bulk citation data for a systematic literature review or meta-analysis workflow
- • Identify the most-cited recent papers on a biomedical topic by combining ESearch and ESummary calls
Not For
- • Retrieving full-text PDFs — PMC only covers open-access articles, not all journals
- • Real-time clinical data or patient records — this is bibliographic metadata only
- • Non-biomedical literature — use Crossref or OpenAlex for broader academic coverage
Interface
Authentication
Completely usable without authentication at 3 requests/second. Registering for a free API key raises the limit to 10 requests/second. Key is passed as a query parameter (?api_key=).
Pricing
NIH-funded public service. No paid tiers exist. Commercial use is permitted.
Agent Metadata
Known Gotchas
- ⚠ ESearch returns a WebEnv/query_key handle that expires — agents must use it within the same session or re-run the search
- ⚠ Large result sets require chained ESearch → EFetch calls; skipping the two-step pattern causes truncated results
- ⚠ The retmax parameter defaults to 20 records; agents that don't set it explicitly miss the bulk of results
- ⚠ XML is the most reliable response format; JSON output is available but occasionally omits fields present in XML
- ⚠ Rate-limit violations return HTTP 429 with no Retry-After header, requiring agents to implement their own backoff
Alternatives
Full Evaluation Report
Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for PubMed / NCBI E-utilities.
Scores are editorial opinions as of 2026-03-06.