# Undermind — AI literature search agent that runs deep searches across 200M+ papers and reads full texts.

> Source: AgentsIndex — https://agentsindex.ai/undermind (structured, researched, re-verified)
> Facts last verified: 2026-10-19

Undermind is an AI literature research service that runs multi-pass deep searches over the Semantic Scholar corpus, reads paper full texts, and writes cited reports in a persistent workspace. It runs in the browser and inside AI assistants through its MCP server.

| Fact | Value |
| --- | --- |
| Website | https://undermind.ai |
| Price | free, then from $20/month billed monthly, $16/month billed annually |
| Free tier | Yes |

Free plan at $0 with standard rate limits on chats and searches; it searches abstracts and metadata, while full-text search and 10x higher usage limits require Pro. The MCP connection and shared workspaces are included.

## What it costs to start

**Free tier** · Yes — A standing Free plan at zero cost includes chat, deep searches, reports, shared workspaces and the agent connection, with standard rate limits.

> Free For exploratory research $0
> — https://www.undermind.ai/#pricing · Pricing section, Free plan card

**Self-serve signup** · Yes — Free, Pro and Team are bought by signing up in the app; only the Enterprise tier routes through Contact Us.

> Sign Up to Purchase
> — https://www.undermind.ai/#pricing · Pricing section, Team plan call to action

**Card to start** · No — The Free plan is a standing zero-cost tier reached by signing up, and a payment method is charged only once a paid subscription starts.

> When you subscribe, your payment method is charged at the start of each billing cycle.
> — https://www.undermind.ai/refund · Opening paragraph of the Refund & Cancellation Policy

**Cancel any time** · Yes — Subscriptions can be canceled from account settings at any time with no fee or minimum term, though the period already paid for is not refunded.

> You may cancel your subscription at any time from your account settings under Manage Subscription.
> — https://www.undermind.ai/refund · Cancellations section

## What it works with

**Public API** · Partly — A programmatic API is sold on the enterprise page with no public reference, while the documented public interface is the MCP server, which has no API-key or service-account path.

> There is no separate API-key or service-account access model for MCP, because Undermind's tools are built around each user's persistent research workspace rather than stateless queries.
> — https://www.undermind.ai/mcp/reference · Connection details, user-scoped paragraph

**MCP server** · Yes — A hosted streamable HTTP MCP server with OAuth and PKCE is documented, listed in the ChatGPT and Claude directories, and included on the Free plan.

> MCP server | `https://mcp.undermind.ai/mcp`
> — https://www.undermind.ai/mcp/reference · Connection details table, MCP server row

## What it does on its own

**Acts on its own** · Yes — The deep search plans its own query sequence, follows citations and key authors across passes, and decides on its own when to stop, with no per-step approval from the user.

> Plans its own searches, follows citations and key authors, and stops only when new searches stop finding relevant papers.
> — https://www.undermind.ai/mcp · Research toolkit, MCP tool list, launch_deep_search entry

**Limits on what it may do** · Partly — Access is user scoped through OAuth with a single scope, so a connected agent gets the whole toolkit or nothing; the only published control is revoking all MCP access afterwards.

> Users can revoke all MCP access from their Undermind account settings.
> — https://www.undermind.ai/mcp/reference · Maintain the connection section

**No code needed** · Yes — The web app needs no engineering work, and assistant connections are one-click listings in the ChatGPT and Claude directories rather than code.

> Teams looking for immediate deployment with no engineering lift.
> — https://www.undermind.ai/enterprise · Undermind for organizations, Web app card, Ideal for list

## What happens to your data

**No training on your data** · Partly — Enterprise agreements promise no training on company data, but the Terms give the vendor a license to use individual users' content in de-identified or aggregated form to develop AI and machine learning systems.

> in de-identified or aggregated form only, developing and improving related technologies, including artificial intelligence and machine learning systems, whether by us or our partners
> — https://www.undermind.ai/terms · Section 3.3 License

**Retention you control** · Partly — Organizations can ask for complete removal and individuals can request deletion, but no setting lets a buyer cap or switch off how long data is kept.

> Request complete removal of your organization's data at any time.
> — https://www.undermind.ai/enterprise · For enterprises section, Own Your IP card

**Data stays in your region** · No — Processing and storage are stated as United States only, with transfers into the US handled by contractual safeguards rather than a region choice.

> We process and store personal data in the United States.
> — https://www.undermind.ai/privacy · International data transfers section

**Open source** · No — The Terms reserve all intellectual property in the service to the company and forbid reverse engineering, and no repository or license is published.

> you acknowledge that all the intellectual property rights, including copyrights, patents, trade marks, and trade secrets, in the Site and its content are owned by Company or Company's suppliers
> — https://www.undermind.ai/terms · Section 2.5 Ownership

**Take your data out** · Partly — Reports and notes are Markdown files and PDF links can be fetched through the agent tools, and privacy rights cover a portable copy of personal data, but no bulk workspace export is documented.

> you may have rights to access, correct, delete, or receive a portable copy of your personal data
> — https://www.undermind.ai/privacy · Your rights and choices section

## Whether a company can buy it

**Security audits** · No — The security overview publishes encryption, access control and third-party penetration testing with an attestation letter under NDA, and names no SOC 2 or ISO 27001 audit anywhere on the site.

> Documentation: Letter of Attestation / pentest report available under NDA.
> — https://www.undermind.ai/security-one-pager.pdf · Contact & Next Steps

**Single sign-on** · Partly — Single sign-on through an existing identity provider is an Enterprise item, described as sitewide organizational login, and is not part of Pro or Team.

> SSO with your existing enterprise IdP
> — https://www.undermind.ai/mcp · Enterprise integration, Works with the agent stack you already run list

**Audit log** · Partly — Organization administrators get an admin dashboard and usage reports on member accounts, which is weaker than a record of who did what and exists only on the top tier.

> may receive reports regarding use of the Site by those Accounts, and may use any other administrative controls the applicable plan provides
> — https://www.undermind.ai/terms · Section 1.4 Organization Administration

**Uptime promise** · Partly — An SLA is negotiated as part of an Enterprise agreement; no uptime figure or public commitment is published for the self-serve plans.

> Custom terms/SLA/security review
> — https://www.undermind.ai/#pricing · Pricing section, Enterprise plan feature list

## Which models it runs on

**Its own model** · No — Undermind builds the retrieval algorithms but runs on third-party model providers it contracts with, rather than training and serving a model of its own.

> Our agreements with our third-party model providers guarantee no long-term data retention and no training for company data.
> — https://www.undermind.ai/security-one-pager.pdf · Key Practices under our Enterprise Agreements, Third-Party Data Protection

**You pick the model** · No — Model assignment follows the plan rather than a user choice: Free gets one class of models and Pro gets another, and no page offers a model picker.

> The latest, most powerful AI models
> — https://www.undermind.ai/#pricing · Pricing section, Pro plan feature list

## A search agent that stops only when a pass finds nothing new

Undermind takes a research goal in prose, asks follow-up questions, then runs what it calls a deep search: semantic search over the corpus, then citation trails, then the key authors, stopping when a pass adds nothing new. The vendor's own illustration of a run shows 340 papers evaluated in the first pass, 186 in the second and 94 in the third, then "Fourth pass added nothing new". A production search [stops on its own after 2.9 minutes on average](https://www.undermind.ai/#benchmark), and the FAQ puts a typical search at about two minutes.

Coverage comes from one source rather than many: "We cover all fields of science using the Semantic Scholar database of over 200 million articles, which contains PubMed, ArXiv, and many other databases." Full texts matter here, and they are tiered. The FAQ states that the [Free tier searches abstracts and metadata](https://www.undermind.ai/#pricing), and that Pro adds compute to search full texts of nearly all open access and preprint articles. Paywalled PDFs are read only after a user uploads them to a workspace.

The output is not a link list. Reports are written into a workspace as Markdown with inline citations back to source papers, searches and starred papers persist, and alerts watch saved topics for new publications.

## The MCP server is the most developed part of the product

Undermind is published as a [ChatGPT app and a Claude connector](https://www.undermind.ai/mcp), with written setup for Cursor, VS Code, Codex, Claude Code, GitHub Copilot, Microsoft 365 Copilot via Copilot Studio, Opencode, Grok, Antigravity and Perplexity. The endpoint is `https://mcp.undermind.ai/mcp` over streamable HTTP.

The tool surface is unusually wide for a research service: deep search and fast search, author and metadata lookup, parallel full-text PDF reading, file writing, starring, and a full folder tree for curating papers. Display tools render cards in clients that support MCP Apps, and fall back to text with links elsewhere.

Authorization is OAuth with PKCE and a single `mcp` scope, and [every request runs as a specific user](https://www.undermind.ai/mcp/reference) inside that user's workspaces. There is no API key or service account: the reference says so plainly, which rules out a shared backend integration that is not acting for one person. Dynamic client registration is not supported, so an unlisted application either hosts a Client ID Metadata Document or emails support to be registered. Connections expire after 90 days without a refresh.

## Two price lists, two billing terms, and one unquantified rate limit

The pricing section carries a control most readers will miss: an Industry and Academic tab beside the monthly and annual switch. On the Academic list Pro is $20 a month, or $16 a month billed annually. On the Industry list the same plan is $75 a month, or $60 billed annually. That is a 3.75x difference for identical features, and no page states who qualifies as academic.

The discount is the vendor's own figure: "You save 20% with annual". Team is priced per person and quoted in blocks of five seats, $18.70 each monthly on the academic list and $75 on the industry list. Enterprise is custom priced and adds increased compute, sitewide organizational login, an admin dashboard and custom terms.

What a plan actually allows is never numbered. Free has "Standard rate limits on chats and searches" and Pro has "10x higher usage limits", a multiple of an undisclosed base, and the [Acceptable Use Policy](https://www.undermind.ai/terms) reserves limits on the number of searches or requests per user. Cancellation is self-serve at any time and access runs to the end of the paid period, but charges are [non-refundable once a billing period has begun](https://www.undermind.ai/refund).

## The training promise depends on which contract you are under

The enterprise material is unambiguous: "Your company's proprietary R&D data stays isolated, is never used to train AI models, and our model providers never retain or train on it."

For an individual account the Terms read differently. Section 3.3 grants the company a license to use User Content for providing the service, for "improving and developing the Services", and "in de-identified or aggregated form only, developing and improving related technologies, including artificial intelligence and machine learning systems, whether by us or our partners". Both statements can be true at once, because an [enterprise agreement takes precedence](https://www.undermind.ai/terms) where it differs, but a researcher reading the no-training card on the marketing page is reading a commitment that attaches to a contract they do not have.

User Content is defined broadly, and includes material retrieved on a user's behalf through integrations, agents and institutional access. The Terms put the rights check on the user: if you lack the rights to an item, do not upload it, and the company states it cannot verify authorization for any source.

## Security is documented, but no audit report is named

The [security overview](https://www.undermind.ai/security-one-pager.pdf) lists encryption at rest and in transit, role-based access control with multi-factor authentication, daily backups with off-site storage, annual staff security training, and manual penetration testing by Oneleet with a certified tester. Evidence is gated: "Letter of Attestation / pentest report available under NDA."

What is absent is a named compliance regime. Neither the one-pager, the enterprise page nor the pricing table mentions SOC 2 or ISO 27001, and the Enterprise tier offers a "security review" rather than a report. Data location is single region: the [privacy policy](https://www.undermind.ai/privacy) states that personal data is processed and stored in the United States, with Standard Contractual Clauses for transfers out of the EEA and UK. Payments run through Stripe. The company behind it is Undermind AI, Inc., registered in Dover, Delaware, founded by two MIT quantum physics PhDs, Joshua Ramette as CEO and Tom Hartke as CTO, and backed by Y Combinator.

## Who it fits, and who should look elsewhere

The buyer this is shaped for is a researcher or an R&D team that needs recall rather than a quick answer, and will accept a few minutes per query to get it. GSK is the reference account, with [1,000 or more researchers using it](https://www.undermind.ai/mcp) inside an internal AI Scientist platform, and the enterprise logo wall adds Novartis, the NIH, Toyota Research and Fraunhofer IKTS.

It fits badly in three cases. A team that wants a stateless query API will find none: access is per user through MCP, and the "Undermind API" is an enterprise sales conversation with no public reference. Anyone whose literature sits behind paywalls does the fetching themselves, since only open access and preprint full texts are read automatically. And a buyer who needs a capacity commitment before signing has nothing to plan against, because no plan publishes a number for its limits.

The Terms also forbid building the product into another offering: the license is for personal or internal business use, and you may not "resell, repackage, or incorporate the Site or its functionality into any product, service, or offering".

## What changed

- 2026-10-19 — Free tier · What it does
- 2026-10-18 — Free tier · Price · What it does

## Keeping it current

These facts are re-checked against Undermind's own pages and updated when they change. If you represent Undermind, sign in and update the listing from your dashboard at https://agentsindex.ai/dashboard. The operations this site accepts from an agent are listed at https://agentsindex.ai/.well-known/agents.json — that catalog needs no key to read, and each entry states its own auth.