API keys, OAuth, and MCP share one workspace-scoped view of patients, encounters, notes, and codes. PHI stays inside the workspace boundary on every path.
Workspace-scoped REST keys generated under Settings, Tools and AI. Drive analytics, internal automation, and custom tools against the same data your clinicians see. Every call carries the workspace boundary, so no key ever reads across tenants.
Standard OAuth 2.0 flow for third-party apps that need delegated access. Users grant explicit consent, scopes are reviewed up front, and tokens carry the same workspace boundary as a key. Use it for partner apps and any tool with its own user surface.
Nextvisit exposes its workspace as a Model Context Protocol server. MCP-aware AI assistants connect directly and read structured clinical data through a typed tool surface, no custom integration build per assistant. Bring your own AI on top of the chart.
Assistants read structured data through typed tools, instead of a one-off integration per product.
Anthropic published MCP, OpenAI adopted it, and most local LLM runtimes support it now. Nextvisit exposes the workspace as an MCP server. An assistant connects once, then reads patients, encounters, notes, and codes through typed tools. The same workspace scope as API keys and OAuth applies.
A clinician can point Claude, ChatGPT, or a local model at the workspace and ask questions against real chart data. Tool calls return structured JSON.
// Assistant invokes a Nextvisit MCP tool
query_encounters({
patient_id: "pat_018f...c1",
range: { from: "2026-01-01", to: "2026-05-01" },
include: ["note", "codes", "peer_review"]
})
// Returns
{
"encounters": [
{
"id": "enc_a1b2",
"date": "2026-04-18T14:32:00Z",
"status": "signed",
"codes": ["F33.1", "F41.1"],
"note": "..."
}
]
}Illustrative shape, not the verified call signature. Read the MCP server guide for the endpoint URL, the live tool list, and connection steps for Claude, ChatGPT, and local LLM clients.
Generate the key under Settings, Tools and AI, then call any documented endpoint.
# List signed encounters for a patient over a date range
curl -sS \
-H "Authorization: Bearer nv_live_8f2c...41a" \
-H "Accept: application/json" \
"https://api.nextvisit.ai/v1/encounters?patient_id=pat_018f...c1&status=signed&from=2026-01-01"
# Response (truncated)
{
"data": [
{
"id": "enc_a1b2",
"signed_at": "2026-04-18T14:42:11Z",
"provider_id": "prv_77a3",
"codes": ["F33.1", "F41.1"]
}
],
"next_cursor": "eyJvZmZzZXQiOjUwfQ"
}Illustrative endpoint and key shape. The live API reference covers pagination, rate limits, and the full resource list.
Pocket and Plaud notes flow in at no license fee.
The heypocket.com hardware recorder posts each recording on completion. AriaMD generates a note in the assigned template, like a live session.
Say the patient name and date of birth in the first minute.
Plaud AI pin and Plaud Note recordings sync into the workspace. Useful if you already carry one for personal capture.
Same convention: name and DOB at the top of the recording.
FHIR R4 and HL7 v2 cover most stacks. Partner integrations cover the rest. Copy-out plus fax handles the long tail. Destination list, bundle layout, and live push state are on the features page.
We run a mock session, draft the note, and show the claim. No slides.