The Register / Research agents / STORM / Co-STORM (Stanford)

STORM / Co-STORM (Stanford)

Name: STORM / Co-STORM (Stanford)
Author: Stanford OVAL lab

STORM (Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking) is an academic system from Stanford's OVAL lab, published at NAACL 2024, that generates an outline and then a long-form cited article on a topic. Co-STORM (EMNLP 2024) adds human-AI collaborative curation with a mind map. It is available as the MIT-licensed knowledge-storm Python package and as a free hosted research preview. The repository has been dormant since September 2025.

storm-project.stanford.edu Source repository Docs Add to compare

In beta. Research preview. The web app at storm.genie.stanford.edu responds (HTTP 200) but the code repository has had no commits since 2025-09-30 and no release since v1.1.0 (2025-01-23); external PRs in 2026 were closed unmerged. Effectively unmaintained code; hosted demo still online.

What reviewers and users praise

Best-evidenced approach to article pre-writing: published evaluation found better-organized, broader articles (Digital Applied summary of the NAACL paper)

What they criticise

Repository dormant since September 2025; users describe it as abandoned
Topic-string input only; a user found it unusable for instructed deep-research tasks

Rating certificate · AW/26-09/64826 Sept 2026

Not enough evidence

STORM / Co-STORM (Stanford)

Stanford OVAL lab · Research agents

Trust BOpen source

Adoption23

Experts—

Crowd—

Experts vs the crowdneeds both scores

Experts—

Crowd—

Price	Free
Use it via	Web app, API
Runs on	web, macos, linux, windows
Licence	MIT
Updated	Evidence collected 26 Sept 2026

How the rating was worked out

Weights for Research agents: adoption 35%, experts 40%, crowd 25%. Fewer than two pillars have enough evidence. The method.

Adoption & momentumweight 35%	23	GitHub stars `32k` → 76 · PyPI monthly downloads `1.8k` → 0 · Reported users `70k` → 23 · `0` releases in 90 days → 0 reach × 0.7 + momentum × 0.3
Expertsindependent reviews · weight 40%	—	Only 1 independent review (3 needed)
Crowdcommunity sample · weight 25%	—	Only 0 community items sampled (6 needed)
Rating	—	Not rated: fewer than two pillars have enough evidence.

Evidence checks. 1 item removed: date does not match the source (1). 1 review from tool directories, AI-written pages or unnamed authors not counted. 1 GitHub issue not counted as sentiment. How evidence is checked.

Trust & safety

Seven dimensions graded A–D from documented facts. Overall: B.

n/e

Permission model

Research/writing system that performs web retrieval only; no code execution. Rubric does not map; not graded.

n/e

Data access scope

Not established.

Data storage

Self-hostable Python package (knowledge-storm); the hosted research preview is optional. source

n/e

Data retention & training

Hosted preview requires accepting terms of service; the terms and data policy could not be read (client-side rendered). Self-hosted use depends on the configured providers. source

Incident history (24 months)

No CVEs found in an NVD keyword search and no security advisories on the GitHub repository. source

Compliance

Academic research prototype; no SOC 2 / ISO 27001 attestation. source

Transparency

MIT-licensed source with published papers and GitHub release notes (last release 2025-01-23). source

The experts

1 named independent reviews found; dated reviews from the last 12 months are scored.

mixed
STORM’s perspective simulation is still the best-evidenced approach to pre-writing — accept that the repository has been quiet for roughly ten months and treat it as a reference implementation.Digital Applied (publication, team byline) · 4 Aug 2026 · digitalapplied.com

The crowd

0 coded posts.

Pricing

As published, checked 26 Sept 2026.

Plan	Price	Notes
Research preview / open source	Free	Hosted preview is free; self-hosted package requires the operator's own LLM and search API keys.

Sources

Every figure above traces to one of these, observed up to 26 Sept 2026.