moonshotclinic.com
Moonshot Clinic AI assistant answering a clinical query

The Chat Box Is the Easy Part

Every major EHR now ships "AI features." Athenahealth has Co-Pilot. eClinicalWorks has Sunoh. DrChrono has ScribeAI. Tebra has AI Notes. They're all variations of the same thing: a chat box bolted onto the side of an interface that wasn't built for it. Drop in a transcription. Get a draft note. Done.

Your AI can summarize the visit you just had. It can't tell you which patients in your panel haven't been seen in 90 days. It can't draft messages to the 23 patients on a specific protocol who are overdue for follow-up labs. It can't pull a P&L by service line. It doesn't know your clinic data — it only knows the document you pasted in.

We built our EHR so the AI lives inside it, with read access to the real clinical data and write access through a draft-then-sign queue that keeps the clinician in the loop. The same ~30 tools that work in the in-app chat work in Claude Desktop too, via our native MCP server. Same data, your AI of choice, one audit log.

The chat box is the easiest 10% of the problem. The other 90% is making the AI actually useful inside a multi-user clinical workflow with real PHI, real billing, real audit obligations. That's what we built.

I tried every EHR's AI feature. They're all the same — a transcription helper bolted onto a system that wasn't designed for AI. You drop in a Quest CSV, you get a draft note, you sign it. That's fine, but it's not what AI can do.

We built Moonshot so the AI can actually see your clinic — your panel, your protocols, your billing. The same tools work whether you're in our chat box or in your own Claude Desktop session via MCP. The data layer is the moat. The chat box is the easy part.

— Tom Kashul, founder of Moonshot Medical & Moonshot Clinic

Your Data, Instantly Accessible

Stop clicking through dashboards. Ask a question. Get an answer.

Natural Language Queries

Type questions in plain English. The AI interprets your intent, queries the database, and returns structured results.

"Which patients are due for labs?"

Find patients overdue for bloodwork, follow-ups, or check-ins. Results include names and direct links to their charts.

"Show me all patients on semaglutide"

Filter by medication, diagnosis, service line, or any clinical attribute. Get a clickable patient list in seconds.

Actionable Patient Lists

Every result links directly to the patient chart. Click to open. No searching, no scrolling, no copying IDs.

Role-Based Data Access

Staff see what their role allows. Providers query clinical data. Front desk queries scheduling. Owners see everything.

Audit Logging

Every AI query is logged with the user, timestamp, and question asked. Full compliance trail for HIPAA audits.

Support Chatbot & Tickets

The same AI powers an in-app support chatbot. Ask how to use any feature, report a bug, or escalate to a human with one click.

AI-Drafted Chart Notes

Ask the AI to draft a SOAP note from an encounter. The draft lands in the clinician review queue and only commits to the patient's chart (encounter_notes) on Sign by a user with edit_charts. Never autonomous.

AI-Drafted Patient Messages

Portal or SMS message drafts queued for review. A clinician with send_messages signs to send. No message leaves the system without a human in the loop.

PDF & Image Attachments

Drop lab PDFs, intake forms, or photos into the chat. The AI extracts and summarizes text; AWS Bedrock vision models read images. Cite back into the chart.

Plug Claude Desktop In via MCP

Same ~30 tools, surfaced through Claude Desktop's Connectors UI in 90 seconds. Use the AI you already use, not a chat box bolted onto the EHR. Read the MCP spec →

Hard Spend Caps

$10/month free AI pool, per-user daily cap, per-tenant monthly hard cap. All configurable in /admin/ai-usage/. Stripe metered billing at 110% of Bedrock cost. No surprise bills.

moonshotclinic.com
AI assistant morning dashboard showing revenue and appointment summaries

Frequently Asked Questions

What kind of questions can I ask?
Anything about your clinic data. Examples: "Which patients are due for labs?", "Show me all patients on semaglutide", "How many new patients did we see last month?", "What's our revenue this quarter?" The AI translates your question into a database query, runs it, and returns results with links to the relevant patient charts.
Is patient data safe when using the AI?
Patient data is protected through multiple defense layers. BAA-covered subprocessors (AWS Bedrock for Claude inference, AWS infrastructure, Stripe, Spruce, AWS End User Messaging). Postgres row-level security enforces tenant isolation at the database layer. Every tool call is audit-logged with the user, payload, and response. No system is fully secure; see our security page for the complete control set.
Can the AI write to the chart or send messages?
Yes, with a human in the loop. The draft_chart_note and draft_patient_message tools propose drafts that land in /admin/ai-drafts/ for a clinician to review and sign. A user with edit_charts signs chart-note drafts; a user with send_messages signs message drafts. There are no autonomous writes — every change passes through a licensed human.
Can I use Claude Desktop or Claude.ai instead of the in-app chat?
Yes. Moonshot Clinic ships a native MCP (Model Context Protocol) server. Connect Claude Desktop or Claude.ai via the Connectors UI in about 90 seconds and you get the same ~30 tools surfaced in the in-app assistant. See /features/mcp/ for the full reference.
How is this different from Athenahealth Co-Pilot or eClinicalWorks Sunoh?
Those tools transcribe your visit and draft a note. That's a useful 10% slice of the problem. Moonshot's AI also has read access to your panel data and write access (with human sign) to chart notes and patient messages. It can answer panel-wide questions, surface patients on specific protocols, and run financial queries — capabilities that require the AI to be deeply integrated with the clinical data layer, not bolted on as a transcription helper.
Is there a limit on how many queries I can run?
Every tenant gets a $10/month AI free pool. Beyond that, usage is billed at 110% of underlying AWS Bedrock cost via Stripe metered billing. Per-user daily caps and a per-tenant monthly hard cap are configurable in the admin panel — no surprise bills.

AI That Works With Your Whole System

Analytics

Dashboards for structured reporting. AI for ad-hoc questions. Use both.

Learn more

Patient Portal

AI queries can surface patients who haven't logged into the portal or completed forms.

Learn more

Scheduling

Ask "Who has appointments tomorrow?" or "Which providers have open slots this week?"

Learn more

See the AI Assistant Live in Your EHR

Book a demo and ask your first question live.

Book a Demo