Patient care navigation and health copilot AI
Your Healthcare Agent
Your Healthcare Agent is a source-available public instruction layer and guide series that helps a patient or caregiver configure Claude as a persistent healthcare navigation agent. It is not a hosted clinical app from the project itself. The agency upside is strong: the spec is explicitly grounded in CAIHL, presumes patient agency, reads bills and EOBs, explains plan type and rights, drafts appeals, supports record access, and helps prepare for appointments and prescription coverage barriers. The main CAIHL cautions are that sensitive health, insurance, and billing materials are uploaded to Claude/Anthropic, full functionality is framed around Claude Pro and Cowork, and the guidance touches legal, insurance, financial, and medical-adjacent domains that require user verification and, for organizational use, qualified human review.
Public-source research has been drafted; final human publication review and change-log detail are still required.
Summary judgment · 86% toward patient-directed
Strongly agency-expanding, with platform and verification caveats
The project gives patients and caregivers a rights-aware agent spec for understanding and challenging healthcare administration, but practical use depends on Claude's consumer platform, careful privacy settings, user verification, and meaningful access to documents, portals, and paid features.
Patient agency
How this tool changes agency
The spec and guides support concrete patient actions: identify plan type and ERISA status, read EOBs and bills, calculate appeal deadlines, draft appeals and portal messages, request records, prepare for appointments, and track claim or refund status across insurer and provider portals.
Use is voluntary, forkable, and copy-based, and the start page says the free Claude version can still read one document at a time. The fuller workflow depends on a Claude account and is framed around Claude Pro, Projects, and Cowork, creating cost, account, region, and platform-dependence limits.
Patient-facing signals
Who does this AI serve?
The official site and spec state that the agent serves the person using it or someone they are caring for, starts from presumed agency, and helps patients exercise coverage, appeal, billing, and records rights. No payer, pharma, provider, or insurer monetization was found in the public sources reviewed.
Can patients tell AI is involved?
AI involvement is explicit throughout: the setup asks users to create a Claude Project, copy a CLAUDE.md instruction block, upload healthcare documents, and optionally use Claude Cowork/browser control. The denial guide warns that Claude is a tool, not a lawyer or doctor.
Can patients meaningfully choose?
Use is voluntary, forkable, and copy-based, and the start page says the free Claude version can still read one document at a time. The fuller workflow depends on a Claude account and is framed around Claude Pro, Projects, and Cowork, creating cost, account, region, and platform-dependence limits.
Can patients correct or challenge what the AI produces?
The spec repeatedly tells the agent to surface options, draft documents, and have the user verify deadlines, amounts, and decisions against authoritative sources. Users can edit prompts and drafts, but there is no public evidence of domain-specific QA, benchmark testing, safety evaluation, or a correction workflow beyond ordinary Claude account controls and GitHub contributions.
Does it help patients understand or act?
The spec and guides support concrete patient actions: identify plan type and ERISA status, read EOBs and bills, calculate appeal deadlines, draft appeals and portal messages, request records, prepare for appointments, and track claim or refund status across insurer and provider portals.
Text findings
Conflict of interest
Related to an already disclosed HugoScore collaborator
The project was designed and built by Nick Dawson, whose OwnChart project is already profiled by HugoScore with a maintainer conflict-of-interest disclosure. Your Healthcare Agent also explicitly cites CAIHL and HugoScore. This profile should be read as a draft and would benefit from third-party review.
Who is left out or burdened?
Cost, document, language, literacy, and portal-access burden
The full workflow assumes a Claude account, often Claude Pro, comfort uploading sensitive documents, English-language guidance, access to plan documents and portals, and enough administrative confidence to verify deadlines and submit appeals. The start page notes that the free version can still help one document at a time and that community organizations may run the workflow for clients, but accessibility, multilingual support, low-literacy testing, and caregiver/proxy safeguards were not documented.
What happens to patient data?
Instruction site is lightweight; patient data flows to Claude/Anthropic and connected services
The reviewed Your Healthcare Agent pages publish static instructions and guides and do not appear to collect health documents directly. Once a user follows the guide, insurance cards, bills, EOBs, denial letters, plan documents, medical records, chats, and portal actions become Inputs, Outputs, Materials, or Actions in Claude. Anthropic's current policy says consumer Inputs and Outputs may be used to train models unless the user opts out, and third-party integrations or services may receive data when Claude acts on the user's behalf.
Are the clinical boundaries clear?
Clear disclaimers, but high-stakes domains require verification
The site says the guides are general educational material, not legal, financial, or medical advice. The spec tells the agent not to provide diagnosis or clinical interpretation and to direct users to verify high-stakes legal or financial matters with authoritative sources. Anthropic's Usage Policy treats healthcare, legal, insurance, and finance guidance as high-risk use cases when advice or recommendations directly affect people, requiring disclosure and qualified human review for those deployments.
Who defined what good looks like?
Maintainer-defined and source-available, with anecdotal case evidence
The behavioral spec and guides are maintainer-defined and source-available, and the public denial guide describes a worked case with a large refund after appeal. Public evidence did not show independent outcome validation, safety testing, legal review, medical review, privacy audit, accessibility testing, or patient-partnered evaluation beyond the project's CAIHL framing.
Review method
Deep public-source review of yourhealthcareagent.org home, setup, instructions, about, guides, raw agent spec, GitHub README and repository metadata, Anthropic Privacy Policy, Consumer Terms, Consumer Health Data Privacy Policy, Usage Policy, and Claude Projects help documentation; no hands-on Claude Project setup, Cowork portal testing, security/privacy audit, legal review, medical review, or outcome validation.
Draft profile · Medium draft, source-available, official sources plus platform policy review