AI research agents
Agents that do the reading for you: they search, screen sources, extract findings and write a report with citations.
By weekly package downloads, the leaders are Kosmos (13.1K), GPT Researcher (9,793), Feynman (2,874). By GitHub stars, the leaders are Autoresearch (100K), BettaFish (42.8K), STORM (31.9K). As of Nov 21, 2026.
-
71
432/4 — — 42.8K 11mo ago 42.8K GitHub starsGid 43 · 2/4 signals Free (OSS)
-
72
432/4 — — 5,493 — 5,493 GitHub starsGid 43 · 2/4 signals Free (OSS)Google Research · Research
PaperOrchestra is a multi-agent framework for automated AI research paper writing.
-
84
41 — — 100K — 100K GitHub starsGid 41 Free (OSS)karpathy · Research · autonomous ML research agent
Andrej Karpathy's experiment in letting an AI agent do ML research on its own: the agent edits a small GPT training script, trains for a fixed five minutes on one GPU, keeps a change only if validation bits per byte improve, and repeats overnight.
-
88
402/4 — — 14.9K — 14.9K GitHub starsGid 40 · 2/4 signals Free (OSS)
-
93
402/4 — — 28K 59d ago 28K GitHub starsGid 40 · 2/4 signals Free (OSS)
-
155
32 9,793 down 13.2% 30.8K 12d ago 9,793 downloads a week down 13.2% 30dGid 32 Free (OSS)Assaf Elovic and contributors · Research · open-source deep research agent
An open-source agent that plans research, gathers and filters many web or local sources and writes a cited report, with support for multiple model providers.
-
205
242/4 13.1K down 9.3% — — 13.1K downloads a week down 9.3% 30dGid 24 · 2/4 signals Free + $200/mo
-
208
23 2,874 — 10.1K 5d ago 2,874 downloads a weekGid 23 Free (OSS)Companion · Research · research agent CLI
Feynman is a command-line research agent that searches academic databases and the web, reads papers and local PDFs, and runs workflows for literature reviews, deep research, paper audits and drafts with cited sources. It runs on the Pi agent with researcher, verifier, reviewer and writer agents.
-
216
22 404 down 3.5% 31.9K 22mo ago 404 downloads a week down 3.5% 30dGid 22 Free (OSS)
-
229
20 — — 5,556 yesterday 5,556 GitHub starsGid 20 Free (OSS)AIPOCH · Research
The open-source AI research workbench for scientific research and agent workflows. Local-first, model-agnostic desktop app with extensible skills, MCP tools and connectors, Python/R execution and traceable artifacts for reproducible research on macOS, Windows and Linux.
-
243
16 — — 5,533 — 5,533 GitHub starsGid 16 FreePraxistNewSapient Intelligence · Research · autonomous research agent system
Praxist runs persistent research loops on projects with a measurable objective, coordinating parallel research agents, task-owned evaluation, evidence retention and synthesis across generations. It is installed via pip and operated through Codex or Claude Code.
-
—
— — — — — Website rank #17.6K (Tranco top-sites list) Free + $20/moConsensus · Research · research agent (scientific literature reviews)
AI search engine for peer-reviewed scientific literature. Its Deep Search mode breaks a research question into a multi-step search plan, runs a series of targeted searches, screens a large set of papers and writes a cited report structured like a literature review, with visuals on scientific agreement, key authors, key claims and research gaps. Other tools include Ask Paper, the Consensus Meter, study snapshots, citation export and collections. It runs in the web browser.
-
—
— — — — — Website rank #35.4K (Tranco top-sites list) Free + $39/mo yearlyElicit · Research · research agent (literature search, systematic reviews, reports)
An AI research platform that searches academic papers and runs a research agent, automated research reports and systematic literature review workflows (screening and data extraction). It spun out of the nonprofit Ought.
-
—
— — — — — Website rank #14.2K (Tranco top-sites list) Free + $20/moPubGenius Inc. · Research · research agent (literature review, systematic reviews)
SciSpace (formerly Typeset) is an AI research workspace whose agent runs multi-step literature reviews, systematic-review screening, data extraction from papers and database searches.
-
—
— — — — — Website rank #333K (Tranco top-sites list) Free + $16/mo yearlyUndermind · Research · research agent (deep literature search)
AI research assistant for scientific literature search, aimed at academic and industry researchers. After asking follow-up questions about the research need, it reads and evaluates hundreds of papers and follows citation trails over several minutes, then returns a report with in-line citations and sortable relevance tables, with a chat for discussing the results. It can monitor topics and alert users to new papers, and can be connected to agents such as Claude and ChatGPT.
No agents match these filters.
“#” is the position among products of the same kind (agents, frameworks, tooling) by Gid Score, our 0–100 composite of adoption, community, attention and momentum. Use the filters and sorting above, or switch to the table view. Methodology · All AI agents.
Two kinds of research agent
- Academic research agents (Elicit, Consensus, SciSpace, Undermind, Kosmos) search scholarly databases, screen papers, extract data into tables and write literature reviews. Some support systematic-review workflows.
- Open-source deep research agents (GPT Researcher, STORM) run on your own machine with your choice of model and search provider and write reports from web or local sources.
General assistants also have research modes; see AI agents for research.
Using them well
A research agent saves the gathering step, not the checking step. Open the citations for every claim you will rely on, ask what the sources disagree about, and say which kinds of evidence you trust.
Head-to-head comparisons
Related guides
Questions about AI research agents
What is the best AI agent for research?
For academic literature, look at the agents built on scholarly databases, which cite papers rather than web pages. For general web research, deep-research modes of general assistants and the open-source agents here work well. By Gid Score, our composite of the public signals tracked here, BettaFish currently ranks first.
Can AI write a literature review?
Several agents here draft literature reviews with citations and extract data from papers into tables. Treat the result as a first draft: check that each cited paper says what the review claims.