Search, filter, and visualize millions of adverse event reports. Ask questions in plain English and get grounded, data-backed answers — no query language required.
Built on the public VAERS dataset co-managed by the CDC and FDA.
1,906,834
Top vaccine types
Total reports
1,906,834
Serious
179,508
Hospitalizations
132,373
Deaths
25,303
Four ways to work with VAERS, from point-and-click filtering to a fully autonomous AI agent.
Search and filter reports by vaccine, symptom, age, state, date, and outcome.
Open
Interactive breakdowns and trends to uncover patterns across the dataset.
Open
Turn a plain-English question into a runnable Elasticsearch query.
Open
An agent that plans, queries the data, and answers your question directly.
Open
Received date across the full dataset
Across all reports
The AI Assistant runs a transparent plan → execute → assess loop. It queries the real dataset at every step, shows its work, and answers only from the numbers it retrieved.
It breaks your question into a short checklist of steps before touching the data.
Each step runs a real query against the VAERS index — counts, breakdowns, or trends.
It checks whether the results answer you, then continues, asks, or writes the answer.
When a question is ambiguous it asks a clarifying question first, and if a task is too large it summarizes what it found and offers to keep investigating.
A public Model Context Protocol server exposes the same VAERS tools to Claude, Cursor, and any MCP-compatible client. Bring your own key, respect the rate limits, and query live data straight from your assistant.
https://vaers.leafnlp.org/api/mcpAdd this URL as a custom connector in Claude, or point any MCP client at it. See the docs for client-specific setup.