Changelog

Over the years, we’ve built Elicit to automate evidence synthesis. With the Elicit Research Agent, you can now take the next step and apply Elicit to high-stakes decisions.

The Research Agent is an AI environment built for high-stakes decisions. It synthesizes evidence across not only scientific literature but a breadth of data sources such as trials, patents, regulatory databases, structured biology databases, the web, and your own uploaded data. Additionally it transforms the evidence into fully cited outputs like reports, data analyses, tables, visualizations, slides, and documents.

We built the Research Agent around these capabilities:

A harness with reasoning principles

Models are trained on BioDecisionBench, a benchmark designed to evaluate reasoning quality and handle uncertainty around decisions in pharma.

Diverse data sources

Make decisions based on the full picture by synthesizing evidence across scientific literature, clinical trials, patents, regulatory and commercial sources, the web, and your own internal data.

Decision-ready outputs

Move from research to decisions in a single workflow by turning evidence into fully cited reports, tables, analyses, visualizations, slides, and documents.

Transparency by design

Trust and defend every decision with transparent methods trail and sentence-level citations.

Knowledge that compounds over time

Improve each decision with Projects that keeps sources, context, instructions, and outputs connected in a single knowledge base.

One of the most important new things we did was create BioDecisionBench, a benchmark specifically evaluating LLMs on reasoning failures in high-stakes pharma decisions. On this benchmark, Elicit highlights more key considerations for decisions (76.7% coverage in Elicit Smartest mode vs. 68.8% for Claude Opus 5 Max). Read the full evaluation blog: https://elicit.com/blog/biodecisionbench-evals.

Early customers are already using Research Agent to prioritize indications, shape clinical strategy, design trials, develop laboratory methods, inform market entry, and strengthen patent strategy. Read the launch post and see examples of Research Agent in action: https://elicit.com/blog/introducing-elicit-research-agent

You can use the Research Agent today at elicit.com or access it through API (https://docs.elicit.com) to bring its capabilities into your own workflows.

Over the years, we’ve built Elicit to be rigorous and systematic for evidence synthesis. We previously released our API in preview to give you access to the same building blocks that make Elicit robust. We've since upgraded our API and today, we're publicly launching it for you to use in your own apps and workflows.
​

We're launching three endpoints to help you in your work:

Search API: This API allows you to search 138 million academic papers and 545K clinical trials using either semantic or keyword search. Our API had the best search results (paper recall) when compared against four of the most popular search tools on BioASQ benchmark.

Reports API: This allows you to submit a research question and generate the answer as a fully cited report based on evidence from hundreds of papers or clinical trials.

Systematic Review API: The most powerful of the three APIs is the Systematic Review API. This API lets you run a full systematic review end-to-end while giving you control over every step across search strategy, screening criteria, data extraction parameters, and reporting.

Together, these let you put rigorous, fully cited evidence directly into your own workflows and agents. See the demo here https://youtu.be/JFRrlRKXqtU.

All these end points are available in the MCP server as well, so Elicit can return cited answers directly in Claude, ChatGPT, Gemini, or any MCP-compatible client.

Early users are already building agentic systems on top of the API. Read our launch post to learn more.

To get started with the API, generate a key at elicit.com/developer. For MCP connections, you can add it as a Claude connector, a ChatGPT plugin, or you can follow these instructions for other clients. The API and MCP server are available on Pro or higher plans.