How can doctors use Claude for clinical questions?
Claude is a strong tool for the writing and reasoning around a clinical decision. On its own it is not a cited clinical reference, and a connector closes that gap. To get cited clinical answers inside Claude, open EvidenceMD's listing in Claude's Connectors Directory at claude.ai/directory/evidencemd, click Connect to Claude and sign in with your EvidenceMD account. From then on, Claude sends clinical questions to EvidenceMD's clinical_reasoning tool, which returns a complete answer with every claim cited as a numbered link to a clinical guideline, clinical journal or peer-reviewed article, framed to your country, specialty and language [1][2]. Add Anthropic's PubMed connector when you want to read the primary papers yourself [10][11]. On patient data, only a HIPAA-ready Claude Enterprise plan with the BAA accepted can hold protected health information. Claude Free, Pro, Max and Team cannot enable HIPAA, and data Claude sends to third-party connectors is outside Anthropic's BAA [12][13]. So write clinical questions without identifiers unless your organisation's agreements cover every tool involved.
Key takeaways
- Claude without a connector writes from training. It is fluent and often right, and a reference it gives is from memory, so open it before you act on it. A connector makes Claude answer from a source it fetched for the question [5][17].
- EvidenceMD is the connector that returns a finished clinical answer. It is listed in Claude's Connectors Directory, connects in one click from the listing, and returns a complete, cited answer from EvidenceMD's clinical reasoning engine instead of a list of papers [1][2].
- Anthropic's healthcare connectors return records, not reasoning. PubMed returns papers; CMS Coverage, ICD-10 and the NPI Registry return coverage policies, codes and provider records. They are the right tools for those jobs, and Claude still does the clinical reasoning itself [8][9][10].
- HIPAA is an Enterprise question. Only a HIPAA-ready Claude Enterprise organisation, with its Primary Owner accepting the BAA, can process PHI. Free, Pro, Max and Team plans cannot enable HIPAA at all [12].
- Connectors sit outside Anthropic's BAA. Even on HIPAA-ready Enterprise, data Claude sends to a third-party connector is not covered by Anthropic's BAA, so the connector needs its own agreement before identifiers go in [13].
- Directory connectors work everywhere Claude runs. One listing serves claude.ai, Claude Desktop, the mobile apps, Cowork and Claude Code, and directory connectors can be suggested in a chat when they fit the task [5].
- Community and Verified labels describe review depth, not capability. Anthropic screens Community connectors before listing and tests Verified ones more deeply; verification is not a security audit. EvidenceMD's listing is new and carries the Community label [2][5].
Why is EvidenceMD the first connector a doctor should add to Claude?
Every other medical connector in Claude hands it material to reason from: papers, codes or registry rows. EvidenceMD hands Claude the reasoning, finished and cited. Five reasons it comes first for a doctor's clinical question, and the caveat that follows them.
It turns Claude into a cited clinical reference
Ask Claude a clinical question with EvidenceMD connected and Claude calls the clinical_reasoning tool, which runs the question through the same Expert clinical reasoning engine that powers evidencemd.ai and returns a complete answer: the assessment, the evidence, the recommendation and the next step [1]. Claude then presents that answer in the conversation. The engine is fine-tuned on clinical reasoning across 40+ specialties and publishes 54.6% on HealthBench Hard for its model, self-reported [3].
Every claim is cited at the claim
Citations arrive as numbered links at the sentence they support, pointing to the clinical guideline, clinical journal or peer-reviewed article it came from, so you can open each one without leaving Claude [1]. The engine retrieves from 40 million+ peer-reviewed papers and clinical guidelines before it writes, so each citation is the source the sentence was written from [3].
It answers for the country you practise in
EvidenceMD reads the country in your profile and frames the country-specific parts of the answer (first-line agents, licensed indications, dose bands, screening intervals) to the guidelines that set practice there, and Claude can pass your specialty and preferred language with the question [1]. A GP in Leeds and a hospitalist in Houston asking the same question get the parts that should differ, differently.
One click from Claude's directory, everywhere Claude runs
EvidenceMD is listed in Claude's Connectors Directory: open the listing, click Connect to Claude, sign in, and it is available in claude.ai, Claude Desktop, the Claude mobile apps and Claude Code, switched on per chat from the + menu [2][5][6]. On Claude Team, a member without permission sees a Request button that sends it to the organisation's admins [5]. Connecting is free on every Claude plan; questions count toward your EvidenceMD plan, failed answers are never counted, and every question is saved to your EvidenceMD history [1][4].
It reasons through drug-to-drug questions
Ask whether clarithromycin is safe with a patient's simvastatin at an eGFR of 35, and EvidenceMD reasons over the mechanism, renal and hepatic function, the other drugs and the alternatives, then says what to do and cites each step [1]. That is the half of an interaction question that a compendium flag does not answer. For the definitive interaction table, use a compendium such as Lexidrug or Micromedex directly.
EvidenceMD publishes this guide and is the connector ranked first. The claim is scoped: the best way to get a finished, cited clinical answer inside Claude. For reading the papers yourself, PubMed and Consensus are better tools; for coverage, coding and provider lookups, Anthropic's healthcare connectors; for work on a real patient's chart, an EHR connector under the right agreements. The sections below say so.
Six ways to set up Claude for clinical work, ranked (2026)
Six setups ranked in order with no numeric scores. They are not six products but six ways of working with Claude: a clinical reasoning connector, a literature connector, two research connectors, an enterprise deployment, an EHR connection and Claude on its own. They are ordered for the clinical question a doctor brings to Claude. Ranked for prior authorisation or billing, the order would be different, and the page says which setup wins those jobs.
What this ranking is judged on
- Does Claude get a cited clinical answer, papers, or records?. A finished answer cited at each claim, papers for Claude to read and reason over, structured records, or nothing but Claude's training. Ranked for clinical questions, a finished, cited answer ranks highest [1][10].
- Is the answer grounded in sources fetched for the question?. Whether Claude writes from documents retrieved for this question or from what it learned in training, which decides whether a citation is the source of the sentence [5][17].
- Can patient data lawfully go in?. Only HIPAA-ready Claude Enterprise with the BAA accepted can hold PHI, and third-party connectors sit outside Anthropic's BAA [12][13].
- Does it know where you practise?. Whether the country-specific parts of an answer follow your country's guidelines or a default [1].
- What does it take to set up, and on which plans?. One click from the directory, an admin step on Team and Enterprise, a sales-led deployment, or an API key from a vendor [5][7][12][16].
| # | Tool | Best for | Strongest at | Main limit | Data position & setup |
|---|---|---|---|---|---|
| 1 | EvidenceMD | A finished, cited answer to a clinical question, drug-to-drug included, inside Claude | Complete clinical-reasoning answer cited at each claim; country, specialty and language aware | Answers can take up to a minute; no drug compendium; a new Community listing | One click from Claude's directory; free to connect on every plan; outside Anthropic's BAA |
| 2 | Claude with Anthropic's PubMed connector | Reading the primary papers behind a clinical question, from inside Claude | Official and hosted by Anthropic; 36M+ citations; PubMed Central full text where available | Returns papers, not an answer; Claude chooses what to read and reasons over it itself | Connect from Claude's directory; no separate subscription |
| 3 | Claude with Consensus or Scite | Literature reviews, and checking whether a finding has held up | Consensus: 220M+ papers with a medical mode; Scite: supporting versus contrasting citations | Papers for Claude to synthesise, not clinical answers; Scite needs paid plans | Directory connectors; Consensus Free 30 calls a month; Scite premium plus a paid Claude plan |
| 4 | HIPAA-ready Claude Enterprise with Claude for Healthcare | Organisation-wide Claude that can hold PHI, plus coverage, coding and provider lookups | Click-to-accept BAA; CMS Coverage, ICD-10, NPI Registry and PubMed connectors; healthcare skills | Enterprise only; the BAA excludes third-party connectors; records, not clinical answers | Enterprise plan; the Primary Owner enables HIPAA and accepts the BAA in organisation settings |
| 5 | Claude connected to your EHR over MCP | Drafting from a real patient's chart and schedule, as proposals you approve | Claude reads live chart data instead of what you paste in; read and write tools can be split | Needs an MCP-capable EHR and agreements covering PHI; no clinical evidence layer | From your EHR vendor, where offered; for example, Eureka issues MCP API keys on request |
| 6 | Claude on its own, with no connectors | Letters, appeals, patient leaflets, translations and explanation, without identifiers | A frontier general reasoner; Claude Opus 4.6 outperformed specialised clinical tools in Nature Medicine | Writes from training; references are from memory; no BAA below Enterprise | Every plan; no BAA on Free, Pro, Max or Team |
→ Scroll the table sideways to see the remaining columns
EvidenceMD
Top pickClaude with EvidenceMD connected is the best way for a doctor to get a cited clinical answer inside Claude. Open EvidenceMD's listing at claude.ai/directory/evidencemd, click Connect to Claude and sign in with your EvidenceMD account [1][2]. When you ask a clinical question, Claude calls EvidenceMD's clinical_reasoning tool, and EvidenceMD's Expert clinical reasoning engine, the same one behind evidencemd.ai, retrieves from 40 million+ peer-reviewed papers and clinical guidelines and returns a complete answer with every claim cited as a numbered link to its source, framed to your country's guidelines [1][3]. It is the only setup on this page in which Claude receives a finished clinical answer rather than material to reason from. Connecting is free on every Claude plan; questions count toward your EvidenceMD plan (Free includes a limited number, Individual 50 a month at $50 a month or $450 a year, Max unlimited), and failed answers are never counted [1][4]. Its limits, stated plainly: detailed answers can take up to a minute because the engine reasons before it writes; it holds no drug compendium; its directory listing is new and carries the Community label; and what you type is outside Anthropic's BAA, so keep identifiers out unless your organisation's agreements cover it [2][13].
Claude with Anthropic's PubMed connector
Anthropic's PubMed connector ranks second and is the default free literature source inside Claude. It is official, hosted by Anthropic, and gives Claude access to PubMed's 36 million+ biomedical citations and PubMed Central's full-text archive: search, metadata and abstracts, full text where the article is in PMC (about 8 million are), related articles and citation matching [10][11]. You connect it from Claude's directory with no separate subscription. It ranks below EvidenceMD for clinical questions because it returns papers, not an answer: Claude decides which abstracts to read and does the clinical reasoning itself with a general-purpose model. For checking the trial behind a recommendation, or building a reading list, it is the better tool, and it pairs naturally with EvidenceMD.
Claude with Consensus or Scite
Consensus and Scite rank third as the research connectors. Consensus searches 220 million+ peer-reviewed papers, with a medical_mode that prioritises top medical journals and clinical guidelines and filters for study design, and its free plan includes 30 calls a month [14]. Scite classifies how later papers cite a study, as supporting, contrasting or mentioning, which makes it the connector to use before you rely on one striking trial; it needs a Scite premium subscription and a paid Claude plan [15]. Both are better than EvidenceMD for a literature review, a grant or a journal club. Both rank below it for the question in clinic, because they return papers and leave the clinical reasoning to Claude.
HIPAA-ready Claude Enterprise with Claude for Healthcare
For a health system, HIPAA-ready Claude Enterprise is the only Claude plan that can lawfully hold protected health information. The organisation's Primary Owner enables HIPAA under Data and privacy and accepts Anthropic's BAA, which is click-to-accept; Team, Free, Pro and Max plans cannot enable it [12]. Anthropic launched Claude for Healthcare in January 2026, adding connectors to the CMS Coverage database, ICD-10 and the NPI Registry alongside PubMed, with skills for prior authorisation and other administrative workflows [8][9][16]. For a revenue-cycle team or a prior-authorisation reviewer, this is the right setup, and it ranks first for those jobs. It ranks fourth here for two reasons. It is a deployment an organisation makes, not a choice a doctor can make alone. And its connectors return coverage policies, codes and records rather than a clinical answer. Note too that enabling HIPAA does not bring every feature under the BAA, and data sent to third-party connectors stays outside it [12][13].
Claude connected to your EHR over MCP
Connecting Claude to an EHR over MCP changes what Claude can see: today's schedule, a patient's chart and timeline, open orders, instead of only what you paste into the prompt [16]. Eureka, for example, documents an MCP server with about 37 tools, each split into a read tool and a separate write tool, so a practice can allow Claude to look without letting it change anything [16]. That makes it the strongest setup for chart-aware drafting: the letter of medical necessity for the 2pm patient, using her actual diagnosis history. It ranks fifth for the clinical question because it brings the patient's data into Claude, not the evidence, so the reasoning is still Claude's own. And because it puts PHI in play, it needs an EHR that offers MCP and agreements covering every part of the path [13][16].
Claude on its own, with no connectors
Claude on its own ranks last for the clinical question and first for the words around it. It is a frontier reasoner: in Nature Medicine (June 2026), Claude Opus 4.6 was among the general models that outperformed OpenEvidence and UpToDate Expert AI on medical benchmarks [17]. It is excellent at the letter, the prior-authorisation appeal, the patient leaflet, the translation and the explanation of a decision made elsewhere. But with no connector it answers from training rather than from a source fetched for the question, so a reference it supplies is recalled, not retrieved, and should be opened before anything rests on it. And on Free, Pro, Max and Team there is no BAA, so nothing identifiable goes in [12]. Use it for words, and connect a source for the medicine.
How can a doctor use Claude safely?
Five rules that hold on every Claude plan, with or without connectors.
Know your plan before anything about a patient goes in
Only HIPAA-ready Claude Enterprise, with the Primary Owner having accepted the BAA, can hold PHI. Free, Pro, Max and Team cannot enable HIPAA, whatever their privacy settings [12]. If you cannot name the agreement, there is not one.
Keep identifiers out of connector questions
Data Claude sends to a third-party connector is outside Anthropic's BAA, even on Enterprise [13]. A connector receives the question Claude sends it, not your whole chat. Write it as a clinical question (age band, findings, drugs, eGFR) without names, dates of birth or record numbers.
Check whether Claude used a source
If Claude answers without calling a connector, the answer comes from training. For a question you will act on, ask Claude to use EvidenceMD or PubMed, and check that the connector was actually called before you rely on the citations.
Open the citation on the claim you would act on
Even with a connector, confirm that the cited guideline or paper says what the sentence says. EvidenceMD places each citation at the claim it supports, so you can open the one that matters [1].
Add connectors only from developers you trust
Anthropic screens Community connectors before listing them and tests Verified ones more deeply, and states that verification is not a security audit; once connected, either kind has the access you grant it [5]. Check who runs a connector and read its privacy policy before you connect it.
When is EvidenceMD not the right choice?
A guide that names no losses is advertising. Four situations where EvidenceMD is the wrong thing to connect, and what to use instead.
You are drafting a letter, a prior-authorisation appeal or a patient leaflet
Use Claude on its own, with no identifiers, or on HIPAA-ready Enterprise
Words are what a general frontier model is best at, and none of that needs a clinical evidence connector [17].
You need the papers themselves for a review, a journal club or a reading list
Use the PubMed connector, Consensus or Scite
They return and organise the literature; EvidenceMD returns an answer, not a bibliography [10][14][15].
You need a Medicare coverage determination, an ICD-10 code or a provider's NPI
Use Anthropic's Claude for Healthcare connectors
Coverage, coding and registry lookups are structured data those connectors return directly [8][9].
You need Claude to work from a specific patient's chart
Use an EHR connector under agreements that cover PHI
EvidenceMD answers the clinical question; it does not read your EHR. Chart-aware work needs the EHR's own MCP server [16].
Which tool fits your role?
The right setup depends on your Claude plan, where you practise and what you use Claude for. Five common situations.
Doctor on Claude Free or Pro
Connect EvidenceMD from the directory for clinical questions, and add PubMed when you want the papers. Keep identifiers out of everything: no individual Claude plan carries a BAA [2][10][12].
Doctor in a practice on Claude Team
Open EvidenceMD's listing and click Request if you cannot connect it yourself; it goes to your admins, who can enable it for the organisation. Team plans cannot enable HIPAA, so the identifier rule still applies [5][12].
CMIO or practice manager
HIPAA-ready Enterprise if Claude will touch PHI, the Claude for Healthcare connectors for coverage and coding, and a written rule that clinical questions sent to any third-party connector are de-identified unless that connector is under its own agreement [8][12][13].
Frequently asked questions
Can doctors use Claude?
Yes, for the work around a clinical decision, and with a connector for the decision itself. Claude is a strong general reasoner and excellent for letters, appeals, patient leaflets and explanation. On its own it answers from training, so connect a source such as EvidenceMD from Claude's Connectors Directory for cited clinical answers. Keep patient identifiers out unless your organisation is on HIPAA-ready Claude Enterprise with the BAA accepted, and every connector involved is covered too.
How do I get Claude to cite medical guidelines and studies?
Connect a source. With EvidenceMD connected from Claude's Connectors Directory, Claude sends your clinical question to EvidenceMD, which returns a complete answer with every claim cited as a numbered link to a clinical guideline, clinical journal or peer-reviewed article. With Anthropic's PubMed connector, Claude searches PubMed and cites the papers it reads. Without a connector, any reference Claude gives comes from memory and should be opened before you rely on it.
What is the best Claude connector for doctors?
For clinical questions, EvidenceMD: it is the connector in Claude's directory that returns a finished, cited clinical answer, framed to your country, specialty and language, rather than papers or records. For reading the literature, add Anthropic's PubMed connector; for Medicare coverage, ICD-10 codes and NPI lookups, Anthropic's Claude for Healthcare connectors.
How do I add EvidenceMD to Claude?
Open EvidenceMD's listing in Claude's Connectors Directory at claude.ai/directory/evidencemd, or search for EvidenceMD under Customize, Connectors. Click Connect to Claude, then sign in with your EvidenceMD account (Google, or email and password). In any chat, turn EvidenceMD on from the + menu under Connectors and ask your clinical question. On Claude Team or Enterprise, an owner adds it for the organisation, and Team members can click Request.
Is Claude HIPAA compliant?
Only on a HIPAA-ready Claude Enterprise plan, after the organisation's Primary Owner enables HIPAA under Data and privacy and accepts Anthropic's BAA. Claude Free, Pro, Max and Team plans cannot enable HIPAA. Even on HIPAA-ready Enterprise, not every feature is covered, and data Claude sends to third-party connectors is outside Anthropic's BAA. HIPAA compliance is a property of the whole arrangement, not of the model.
Can I put patient information into Claude Pro or Max?
No. Claude Pro and Max are individual plans that cannot enable HIPAA or carry a BAA, whatever privacy settings are turned on. De-identify everything: remove names, dates of birth, record numbers and exact dates before pasting, or use your organisation's HIPAA-ready Enterprise deployment.
What is Claude for Healthcare?
Anthropic's healthcare offering, launched in January 2026: HIPAA-ready Claude for organisations on Enterprise plans, connectors to the CMS Coverage database, ICD-10 and the NPI Registry alongside PubMed, and skills for workflows such as prior authorisation. Its connectors return coverage policies, codes and provider records. They are built for administrative and research work, not for returning a cited clinical answer.
Does Claude have a PubMed connector?
Yes. Anthropic hosts an official PubMed connector that gives Claude PubMed's 36 million+ citations and full text from PubMed Central where available. Connect it from Claude's Connectors Directory; it needs no separate subscription. It returns papers for Claude to read, so pair it with EvidenceMD when you want a finished clinical answer.
Can I use OpenEvidence inside Claude?
Not through an official connector, as of October 2026. Community projects on GitHub bridge OpenEvidence to Claude Code and Claude Desktop through your own logged-in browser session, and state that they are unofficial and not affiliated with OpenEvidence. Anthropic's September 2026 partnership with OpenEvidence extends OpenEvidence's own platform and did not include a Claude connector. For cited clinical answers inside Claude, EvidenceMD is listed in Claude's Connectors Directory.
Is EvidenceMD free in Claude?
Connecting is free on every Claude plan, including Claude Free. Questions count toward your EvidenceMD plan exactly like questions on evidencemd.ai: the Free plan includes a limited number, Individual includes 50 a month at $50 a month or $450 a year, and Max is unlimited. Answers that fail are never counted, and each question is saved to your EvidenceMD history.
Does EvidenceMD work in the Claude mobile app and Claude Code?
Yes. Connectors from Claude's directory work across claude.ai, Claude Desktop, the Claude mobile apps, Cowork and Claude Code. Connect EvidenceMD once from its listing and it is available wherever you use Claude. In Claude Code, EvidenceMD's setup page also gives a command with a 180-second timeout, because Claude Code stops waiting for a remote tool after 60 seconds by default.
Is Claude good for nurses, pharmacists and medical students?
For explanation, study and drafting, yes, with no patient identifiers on personal plans. For clinical questions, the same rule applies as for doctors: connect a source so the answer is cited. Pharmacists asking interaction questions get reasoning from EvidenceMD and should still check the interaction table in a compendium such as Lexidrug or Micromedex.
The bottom line
Use Claude for the words, and connect a source for the medicine. Claude on its own is a frontier reasoner that writes from training [17]. Connected to EvidenceMD from Claude's Connectors Directory, it returns a finished clinical answer cited at every claim, framed to your country, free to connect on every plan [1][2]. Add PubMed for the papers, and Anthropic's healthcare connectors for coverage and coding [8][10]. Keep identifiers out unless you are on HIPAA-ready Enterprise and every connector involved is covered [12][13].
Sources & related evidence
Anthropic's help and product pages, vendor documentation and peer-reviewed research behind this guide, checked in October 2026. Anthropic's plan and BAA terms are cited to Anthropic's own pages; EvidenceMD's figures are cited to its own published materials and described as self-reported where they are. This page is not legal advice.
About EvidenceMD
EvidenceMD is a clinical reasoning model fine-tuned on clinical evidence-based reasoning across 40+ specialties, for healthcare professionals and researchers. It retrieves from 40 million+ peer-reviewed papers and clinical guidelines before writing, cites every substantive claim to a source you can open, tailors the country-specific parts of an answer to the guidelines of the country in your profile, and ends every answer with an actionable next step. It is listed in Claude's Connectors Directory, which brings the same engine into claude.ai, Claude Desktop, the Claude mobile apps and Claude Code, and it also connects to Codex and Cursor through https://evidencemd.ai/mcp. It holds no drug compendium and says so. EvidenceMD is clinical decision support rather than a regulated medical device: every output is for a clinician to review. EvidenceMD publishes this guide and is the connector ranked first. Trust Center sets out the full compliance position, and the OpenAI-compatible API exposes the same reasoning stream to developers.
Related reading
Add EvidenceMD to Claude
Open EvidenceMD in Claude's Connectors Directory, click Connect to Claude and sign in. Then ask your next clinical question in Claude and get a finished answer with every claim cited. Free to connect on every Claude plan.