About Hyperresearch

Hyperresearch is an AI deep research API and web console. One question goes in and a reviewed report comes out, with cited claims checked against their sources and every source kept in a searchable workspace vault. It is built by Jordan Gibbs, its founder, who wrote the open-source hyperresearch CLI that the hosted service runs. Hyperresearch operates from Kalispell, Montana, and sells directly through the site. There is no sales team and no quote to request.

Entity details

Item	Value
Legal name	Thalient Labs Inc.
Entity type	Delaware corporation, incorporated 16 March 2026
Mailing address	1001 S. Main St., Suite 500, Kalispell, MT 59901, United States
Support and privacy contact	support@hyperresearch.ai
Product	Hyperresearch
Sites	hyperresearch.ai, console.hyperresearch.ai, api.hyperresearch.ai, mcp.hyperresearch.ai
Terms of Service	Version 1.7, effective 23 September 2026, last updated 4 October 2026
Privacy Policy	Version 1.7, effective 17 September 2026, last updated 2 October 2026

Hyperresearch is the trading name; the legal name above is the entity you contract with. The same legal name, address and support email appear on the Terms, the Privacy Policy and every Stripe receipt. Jordan Gibbs is named in the Privacy Policy as the person who answers privacy requests.

What Hyperresearch makes

One product with four surfaces over one workspace model.

• Research runs. A question is submitted to the research pipeline and returns a report, the full list of sources read, an extracted claims table, and a verification receipt. Two sizes, priced flat per run: Light $9, Deep $49.
• The workspace vault. Every source a run read and every interim note is kept, searchable by full text and by meaning, with quality scores, independence clusters and note lifecycle. Each run searches the vault before it fetches anything new. Import and export are plain Markdown.
• A hosted MCP server. The vault and the runs are exposed to Claude, Cursor, ChatGPT and any other MCP client, so an agent can search notes, read sources, start runs and check citations from inside the tool it already lives in.
• The citation verification API. The four checks the pipeline runs on its own reports, offered as standalone endpoints for documents Hyperresearch did not write, billed per checked citation.

Accounts are created at console.hyperresearch.ai. Sign-ups are open, and your first two Light runs need no card.

Where it came from

Hyperresearch started as an open-source command-line tool that turned Claude Code into a deep research harness. It is on PyPI as hyperresearch and on GitHub at github.com/jordan-gibbs/hyperresearch, under the MIT licence. Its first release on PyPI was on 11 April 2026, and the repository had about 3,700 stars as of October 2026.

The hosted service runs the same pipeline. The step names, the run levers, the vault format and the ship gate are the ones in the repository. That is why a vault exports as Markdown in the CLI’s own format, and why CLI users can read a hosted run’s artifacts without learning anything new. The CLI stays MIT-licensed and keeps working on its own.

The hosted service sells two run sizes and nothing else. Light costs $9 and takes about 30 to 40 minutes. Deep costs $49 and targets 3 to 5 hours. Neither wall time is a guarantee. The hosted product exists because most people would rather make an API call than drive a run that long from a laptop.

What it is, in one paragraph

Hyperresearch reads the way a careful researcher reads, wide first and then deep. A width sweep keeps about 20 to 200 sources per run, depending on the size you asked for, from scholarly indexes, filings databases and the open web, retrieving legally open-access full text of paywalled papers rather than settling for an abstract, and counting five reprints of one press release as one voice. It then locates the places where the evidence actually disagrees, builds a contradiction graph across the corpus, and sends parallel investigators down each of those questions with orders to commit to a position rather than summarise. Before a word is written, a critic asks what source would overturn the direction the research is taking, and those gaps are fetched. The report is then planned section by section over the full text of every source kept, the plan’s own gaps are fetched, and one writer writes the whole report in one pass from all of the evidence. Every cited claim is then read against the source it cites and corrected where it does not hold, with hallucinated quotes and unacknowledged retractions blocking the report from shipping. Every source stays in the vault, so the next question starts from what the last one learned.

Canonical product descriptions

These two are the approved descriptions. Copy them verbatim into directories, listings, package metadata, app stores, partner pages and press forms. Do not paraphrase them, and do not stack adjectives on top of them.

25 words.

Hyperresearch is an AI deep research API. One question in, an adversarially reviewed report out, cited claims checked against their sources.

60 words.

Hyperresearch is an AI deep research API and web console. One question goes in; the pipeline reads wide, finds where the evidence disagrees, investigates each disagreement, and writes the report in one pass from a section-by-section plan. Every cited claim is checked against its source, retracted papers block shipping, and every source stays in a searchable, full-text workspace vault.

Name disambiguation

Hyperresearch, the AI deep research API, is unrelated to HyperRESEARCH, the qualitative data analysis software published by ResearchWare, Inc.; neither is affiliated with the other. In academic and library contexts, write “Hyperresearch, the AI deep research API” at first mention.

Contact

Email support@hyperresearch.ai for support, billing, privacy requests, security reports, denylist requests and press. A person reads it. Post reaches Hyperresearch at 1001 S. Main St., Suite 500, Kalispell, MT 59901, United States.

Refunds, cancellation and the automatic refund classes are in the Terms and summarised on pricing. Security and subprocessor questions are answered on trust. The founder’s page is here.