Asta

What it does

Asta takes in research questions, natural-language requests, and scientific literature queries, and produces cited literature results, structured summaries, and data analyses. Its literature tools search an abstract index and a full-text index, then return papers with explanations of relevance or synthesized mini-reviews where each claim links to a citation. Its data analysis capability turns natural-language questions into analyses, visualizations, and pattern identification.

Researchers use the Find papers, Summarize literature, and Analyze data functions to navigate large bodies of literature, form hypotheses, and communicate findings. The product emphasizes transparency: outputs are cited and traceable to sources, and the agents and evaluation framework are described as open source. AstaBench, a companion benchmarking suite, and Asta resources, including open-source baseline agents and MCP-accessible tools, support developers building scientific agents.

Asta suits researchers, especially those working across disciplines, and developers building or evaluating scientific AI agents. Access is currently through an Asta Preview sign-up, and the Analyze data function is described as beta for select partners. Asta is described as an evolving initiative that augments rather than replaces scientists.

What it can do

• Literature search that reformulates queries, follows citations, and explains paper relevance
• Summarization of literature into structured, cited mini-reviews over abstract and full-text indexes
• Natural-language data analysis that generates visualizations and identifies patterns
• Open-source agents and baseline agents available to fork and compare
• AstaBench evaluation framework with leaderboards and benchmarks for scientific agents
• Open-weights and post-trained science models, including AstaBrief for cited report generation

Works with

• Model Context Protocol (MCP)
• GitHub

Example workflows

Literature review for a research question

Input: a research question. Steps: Asta reformulates the query, searches for papers, follows citations, and evaluates relevance. Output: a list of papers with explanations of why each is pertinent.

Cited summary of a topic

Input: a complex scientific question. Steps: Asta filters and clusters evidence from abstract and full-text indexes. Output: a structured mini-review where claims link to citations and excerpts.

Before you choose

• Analyze data is described as beta for select partners, so availability may be limited.
• The open-source versions of Paper Finder and Scholar QA may differ from production versions, per the blog.
• Asta is described as experimental in parts (for example, Asta v0 is not used in production), so features are evolving.

Listed under

• Data
• Research
• Agents
• AI research
• Benchmarks
• Developers
• Literature review

Get started

Sign-up is the stated entry point; Analyze data is beta for select partners.

1. Sign up for Asta Preview at the provided link to request access.
2. Explore the Find papers and Summarize literature functions with a research question.
3. For developers, review AstaBench and Asta resources on GitHub to evaluate or fork baseline agents.

Pricing and billing terms for Asta are not stated on the product pages.

We couldn’t find published prices for Asta. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Asta: Advancing Scientific AI with Agents & Benchmarks ↗allenai.org
• Asta Agents: AI Tools for Scientific Research ↗allenai.org
• Asta: Accelerating science through trustworthy agentic AI | Ai2 ↗allenai.org
• Ai2 documentation ↗docs.allenai.org

Company website

allenai.org: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.