STORM (Stanford)
Stanford's open-source research prototype that drafts cited reports
Overview
STORM is Stanford OVAL's open-source research system: it simulates multi-perspective questioning against internet sources, builds an outline, then drafts a long Wikipedia-style article with citations. Co-STORM lets a person steer the research. The code is MIT-licensed and self-hostable, a free hosted preview requires sign-in, and maintenance has slowed since late 2025.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 12 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
What you can produce with STORM (Stanford)
- A long, cited, Wikipedia-style report from a single topic
- Simulated multi-perspective expert interviews to gather sources
- A structured outline before the full draft
- Co-STORM mode for steering the research conversation yourself
- MIT-licensed knowledge-storm package with your own model and search keys
- A free hosted research preview (sign-in required)
STORM is a clear example of structure-first long-form generation: it surveys perspectives, simulates source-grounded interviews and outlines before drafting, which in Stanford's evaluation produced better-organised and broader articles than a single-pass retrieval baseline. Co-STORM lets a person steer.
It is a research artefact, not a service. Self-hosting needs Python and your own model and search keys, nothing has been merged since September 2025, and the hosted preview's terms let inputs be reused for research. Use its output as a cited first draft to edit.
What we found
STORM is a research system from the Stanford Open Virtual Assistant Lab (OVAL) that turns a topic into a long, Wikipedia-style report with citations. Its central idea is to do the pre-writing properly: it discovers different perspectives on a topic, simulates conversations in which a writer asks questions of a topic expert grounded in internet sources, and builds an outline before drafting. The paper by Yijia Shao, Yucheng Jiang, Theodore Kanell, Peter Xu, Omar Khattab and Monica Lam appeared at NAACL 2024, and Wikipedia's newsletter, The Signpost, reviewed it in August 2024.
In the paper's evaluation, more STORM articles were judged well organised, a 25-point absolute increase, and broad in coverage, 10 points, than those from an outline-driven retrieval baseline. The experienced Wikipedia editors involved also identified what remains hard: source bias carried into the article and unrelated facts linked together. The README is equally plain that the output is not publication-ready, so treat it as a sourced first draft.
Co-STORM, published at EMNLP 2024, adds a collaborative mode in which a person can join the research conversation and steer it. Both systems ship in the MIT-licensed knowledge-storm Python package and run with your own language-model and search keys, or on your own documents.
Maintenance is the caveat. The last tagged release is v1.1.0 from January 2025, the last change merged to the main branch is dated 30 September 2025, and outside contributions keep arriving without being merged; in July 2026 a maintainer closed a batch of them as unreviewable. Earlier versions of this page described the repository as actively maintained. The hosted research preview is still online, requires sign-in, and its terms say inputs may be used for research and redistributed under a Creative Commons licence, so keep confidential material out of it.
STORM suits researchers and writers who want a structured, cited starting point and can run Python. It is not a supported product.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including STORM (Stanford)'s official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with STORM (Stanford) unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to STORM (Stanford) directly →
Spotted something out of date? Suggest an update →
Related reading
AI & ML
DeepSeek Open-Sources DeepInfer, the Engine That Serves Its V4.1 Flash Model on NVIDIA and...
Health & Wellness
Fine-Particle Air Pollution Linked to Higher Hospital Admission Risk for Up to 21 Days, St...
Health & Wellness
A Guide for New Mothers Missed Its Main Goal but Nearly Doubled Primary Care Visits
Alternatives to STORM (Stanford)
More in Research & Data
