EMR MCP connects Claude to your behavioral health EMR — secure, read-only, and signed in as each staff member. Add Salesforce and CallTrackingMetrics, and the whole picture becomes a single question: census, admissions, documentation, and what each marketing source actually costs you per admit.
Every answer is pulled live from your EMR instance and signed per request. No patient data cached, none stored.
Nearly every treatment center runs the same three systems. Your EMR holds the chart. Salesforce holds the pipeline. CallTrackingMetrics holds the phones. Apart, each answers half a question. Connected to Claude, they answer the whole one.
The chart and the census. This connector is our software, built and hosted by us: read-only, permissions mirrored from your own EMR user directory, PHI minimized by default, every call attributable to the staff member who made it. It is what the monthly subscription covers.
Referrals, inquiries, and the admissions pipeline. Salesforce publishes its own MCP server — we configure it against your org and connect it to your Claude workspace as a one-time setup.
Calls, sources, and spend. CTM publishes its own MCP server too — same arrangement. We wire it up once, and then it is simply there when someone asks.
Marketing spend lives in one system and clinical reality lives in another, so the questions that matter most across them usually cannot be asked at all — they become a reporting request that takes a week.
What is ours, and what is not. The EMR connector is our software, and the guarantees on this site — read-only, PHI minimization, EMR-mirrored permissions, per-user audit — describe it. Salesforce and CallTrackingMetrics are their vendors’ own MCP servers: we set them up for you, but they authenticate through those vendors, carry those vendors’ own access model, and are governed by those vendors’ terms. Connecting them is your decision and your risk to assess.
No warehouse, no analytics team, no six-month implementation. Your EMR, Salesforce, and CallTrackingMetrics connected to Claude, and one sentence describing what to show. Build a view like this in minutes — or skip the dashboard and just ask.
| Rep | Inquiries | Admits | Conv. | Speed to lead |
|---|---|---|---|---|
| A. Whitmore | 214 | 78 | 36.4% | 1.4 min |
| L. Barrett | 236 | 81 | 34.3% | 2.1 min |
| D. Cho | 198 | 62 | 31.3% | 2.8 min |
| Clinician | Caseload | Open notes | Missing BPS | Compliance |
|---|---|---|---|---|
| Clinician A Residential | 22 | 14 | 5 | 56.8% |
| Clinician B Detox | 18 | 11 | 4 | 58.3% |
| Clinician C PHP | 20 | 4 | 1 | 87.5% |
Sample dashboard. Layout and metrics are representative of what the connected systems return; the figures and names shown are illustrative, not a report on any real organization or person. What is real is where each number comes from — census and documentation from your EMR, pipeline from Salesforce, call sources and spend from CallTrackingMetrics — and that no single vendor puts them in one view for you.
No IT project. No changes to your EMR configuration. No new software for staff to learn — it's Claude, connected to the system of record.
Your EMR administrator generates API credentials from your own instance — Settings → API management. We work with both API V3 and V4. You stay in control; you can revoke it any time.
access_id · secret_key · app_idWe sign a BAA, validate your credentials live against your EMR, and stand up your organization's private connector on its own subdomain with your sign-in policy.
yourorg.emrmcp.com/mcpAdd the connector in Claude's settings — once, org-wide. Each staff member signs in with their own work account, so access and audit stay per-person.
Settings → Connectors → AddNo write path to your EMR exists in the product. The connector cannot create, change, or delete EMR data — structurally, not by policy.
Your EMR user directory is the access list. Staff sign in with their work account; their EMR roles set what they can ask, their location assignments set the boundaries — and removing a user in your EMR revokes their access at the next nightly sync.
Requests default to the minimum necessary. Detox census counting happens server-side, so headcount questions come back as counts, not patient records.
No patient data is ever cached or stored — it passes through, signed per request. We hold only three things: your staff directory (names, emails, roles, locations) to enforce permissions, an audit record of who ran what, and your API credentials, envelope-encrypted.
Your organization runs on its own subdomain with its own credentials and sign-in policy. Tokens issued for your org are useless anywhere else.
We sign a BAA with every customer, and recommend Claude Enterprise with a BAA signed with Anthropic on your side. Every tool call is logged — who, what, when — never patient arguments or results.
Answers land wherever your staff already work — Claude desktop, web, or mobile — through the MCP connector. Because EMR data then lives in those conversations, we strongly encourage every customer to secure a Claude Enterprise account with a BAA signed directly with Anthropic. That is the ideal configuration, and the one we consider suitable for unrestricted, day-to-day clinical use.
You don’t have to wait on paperwork to get value. Staff placed on our aggregate access tier reach only counts and facility-level rollups — census and occupancy totals, detox headcounts, and which facilities exist. Those answers are computed inside our service and come back as figures, never patient records: only the figures leave it, so no patient record crosses into your Claude workspace. For most operational and executive reporting, that is the whole job.
Chart-level tools — patient records, evaluations, assessments — stay switched off for those users until you say otherwise. It is a per-person setting mirrored from your own EMR roles, so you decide who is limited to statistics and who is not.
Purpose-built tools for the questions treatment centers actually ask — each one scoped to a specific job, signed per request, and read-only.
LocationsFacility list with location IDs — the anchor for every other query.
CensusAdmissions by location and date range, PHI-minimized by default.
OccupancyBeds, capacity, and availability for daily census questions.
Recent updatesWhat changed this week — patients created or updated in a window.
Patient recordOne record by casefile ID, detail flags opt-in only.
AppointmentsScheduler pulls by date range and location.
EvaluationsEvery assessment on a chart — biopsychosocial, ASAM, nursing, progress notes — with author and dates.
Evaluation detailRead a full assessment. Long documents return complete instead of truncated, so nothing clinical is silently lost.
Detox censusVerified in-care headcount that handles pre-dated discharges correctly. Counts only — never records.
Every clinician who needs it gets access — charging per user would only discourage the people the tool is meant to help. You scale by facilities, not headcount.
We configure Salesforce’s own MCP server against your org and connect it to your Claude workspace, so referrals, inquiries, and pipeline become askable alongside the chart.
The same for CTM: calls, sources, and spend, wired up once. This is the piece that makes cost per admission answerable at all.
Setup is one time. After that there is no additional monthly for the connected systems — your subscription covers them. Each vendor’s own licence and terms still apply on their side, and their MCP servers are not covered by the guarantees described for the EMR connector.
All plans include every tool, unlimited staff accounts, per-user audit logging, and a signed BAA. We accept all major credit cards and ACH bank transfer. Card payments carry a 3% processing fee; ACH does not. Billing is prepaid — the full term is invoiced at the start of the period. Enterprise can pay against a PO on net-30 terms. Cancel any time — your connector keeps working through the period you've paid for.
Pricing details & FAQ →No. The product is read-only by construction — there is no code path that writes to your EMR. Your team asks questions; nothing is ever created, updated, or deleted.
API access on your instance — V3 or V4 — and a set of API credentials (Settings → API management). If API access isn't enabled yet, your EMR vendor can turn it on. We validate your credentials live during onboarding.
We sign a BAA with every customer. Requests default to minimum-necessary PHI, aggregate tools return counts instead of records, no patient data is stored, and every call is attributable to a signed-in staff member. On the Claude side, we recommend using Claude Enterprise only, with a BAA signed with Anthropic — EMR answers flow into your team's Claude conversations, so that agreement is a key part of keeping the full workflow HIPAA compliant. We walk you through it during onboarding.
We suggest Claude Enterprise only, with a BAA signed between your organization and Anthropic — that's the configuration we recommend for remaining HIPAA compliant, since EMR data appears in your staff's Claude conversations. An admin adds the connector once for the whole workspace; each user still signs in individually, so access and audit stay per-person.
About 15 minutes once you have your API credentials: we provision your endpoint, you paste it into Claude, staff sign in, and the first question gets answered. No integration project, no changes to your EMR.
$999/month up to 3 facilities, $1,499/month up to 15, and custom pricing beyond that — unlimited staff on every plan, and two months free if you pay annually. There is no free tier; plans are month-to-month and you can cancel any time. Full detail in pricing.
Real questions against your own instance — about 15 minutes from credentials in hand to your first answer.
Book a demo