DEX
Sell your expertise Join the beta

LIVE TODAY

Three experts are live. More on the way.

Each one is an expert for a single field. It answers from real source material, shows you where every answer came from, and says so when it can't. A source on every answer, or it's free. It plugs into your agent as a tool.

c7c19f25050d → "the examiner must present a convincing line of reasoning why a person skilled in the art would combine the references"

5,615 chunks CITE-OR-REFUSE
EVAL ★★★★★ CITE-OR-REFUSE: PASS Passed our trick-question test. Not a user rating.

Patent (MPEP)

updated 2026-06-30
Source: USPTO MPEP (public record)

Examiner procedure from the USPTO's own manual, 8 core chapters. Every answer shows the exact MPEP passage it came from.

$0.10 / cited answer PREMIUM

$0.00 / refusal

"Can an examiner combine two references to reject my claim as obvious?" Read this session's full transcript → How to connect

85ffb56cac4b → "capture actual production traces (or at least UAT-level traces) into your eval datasets"

10,693 chunks CITE-OR-REFUSE
EVAL ★★★★★ CITE-OR-REFUSE: PASS Passed our trick-question test. Not a user rating.

AI Engineer

updated 2026-07-07
Source: public conference talks

Applied AI engineering from AI Engineer conference talks: agents, evals, retrieval, production LLM systems. Every answer shows its source.

$0.02 / cited answer

$0.00 / refusal

"What do practitioners recommend for evaluating LLM agents in production?" Read this session's full transcript → How to connect

117b584cbfc6 → "customers beating a path to your door"

7,649 chunks CITE-OR-REFUSE
EVAL ★★★★★ CITE-OR-REFUSE: PASS Passed our trick-question test. Not a user rating.

YC Startup Library

updated 2026-07-07
Source: public YC startup videos & essays

Startup guidance from YC partner talks and founder lectures: fundraising, product-market fit, growth. Every answer shows its source.

$0.02 / cited answer

$0.00 / refusal

"How do I know if I have hit product-market fit?" Read this session's full transcript → How to connect
Tax COMING SOON

Tax

Planned source: IRS publications (public record)

Get notified →
Immigration COMING SOON

Immigration

Planned source: USCIS policy manual (public record)

Get notified →
Aviation COMING SOON

Aviation

Planned source: FAA regulations (public record)

Get notified →
Securities COMING SOON

Securities

Planned source: SEC rules & guidance (public record)

Get notified →
Workplace safety (OSHA) COMING SOON

Workplace safety (OSHA)

Planned source: OSHA standards (public record)

Get notified →
Environmental (EPA) COMING SOON

Environmental (EPA)

Planned source: EPA regulations (public record)

Get notified →

WHAT IS AN EXPERT PACK?

An expert, for one field, your agent can call.

An expert pack gives your AI agent the knowledge of someone who has spent years in one field. Your agent consults it before it answers, so you get a real domain expert's answer with the exact source attached. Not a confident guess.

01

Your agent asks a question it has to get right.

02

The pack looks for the answer in the field's real source material, and only answers when the source clearly backs it up.

03

It answers with the source attached, or it says it doesn't know. No source, no answer, and a "don't know" costs nothing.

An answer from a real domain expert.

Your agent works from what someone in the field actually knows, so you stop getting confident filler on the questions that matter.

You can check every answer.

Each one shows the exact source it came from, so you can verify it in seconds instead of taking it on faith.

It won't bluff.

When the source can't back an answer, the pack says so instead of making something up. That answer is always free.

HOW IT WORKS

Give your agent a domain expert to call.

DEX is short for domain experts. Here is what happens when your agent needs one.

1

Connect

claude mcp add dex-patent \
  ⟨endpoint issued with your beta key⟩

ENDPOINT FORMAT · YOURS IS ISSUED WITH YOUR BETA KEY

https://mcp.dexforagents.com/{your-pack}

2

Ask

Your agent queries the pack over MCP, mid-task, like any other tool call.

How does the USPTO analyze subject matter eligibility under §101?
3

Cited or free

Every answer returns with a citation and chunk key to the primary source. If the sources can't back an answer, the pack refuses and refusals cost nothing.

chunks ec4dd92903fb +2 · GROUNDED · 8.4s · $0.10
REFUSED · 4.3s · $0.00

MCP is the one-line way agents plug into tools. Claude Code, and most agent frameworks, already speak it.

Endpoints are issued per beta batch. There is no public URL yet.

Get a beta key

WHERE YOUR QUERIES GO

Your agent's queries go to the pack server, which retrieves from the corpus and generates the answer through Anthropic's API. We log queries, retrieved chunks, and answers to debug retrieval and run evals. We don't use your data to train models and we don't sell or share it.

TESTED CLIENT

Works with Claude Code. MCP is an open protocol; other clients are untested until we say otherwise. Beta batches are onboarded by hand, so bring the client you actually use.

Come with a question your agent keeps getting wrong.

Join the beta