The healthcare MCP server for AI agents

Arlo Health runs a Model Context Protocol server at https://mcp.arlohealth.ai. Any MCP client can connect over streamable HTTP with OAuth 2.1 and act in healthcare for its user. The agent gathers the story and keeps the thread moving. A licensed clinician does the part that needs a license. Control returns to the agent.

It works in Claude, ChatGPT, OpenClaw, Claude Code, and custom agent runtimes. One integration covers consultations, prescriptions, labs, referrals, health records, care requests, and care prices.

Server details

• Server URL: https://mcp.arlohealth.ai
• Transport: Streamable HTTP
• Authentication: OAuth 2.1 with PKCE. Connector hosts run the flow in their own UI.
• Server card: https://mcp.arlohealth.ai/.well-known/mcp.json
• Agent guide: https://mcp.arlohealth.ai/AGENTS.md and https://mcp.arlohealth.ai/llms.txt
• Cost: Free to integrate. The patient approves each clinician visit on a Stripe link, and nothing is charged until a clinician joins. The care pricing tools are free and never charge anyone.

What an agent can do

• Consultations. Start a conversation with a first-person account of the problem, answer AI triage questions, and hand the patient to a licensed clinician who can diagnose, prescribe non-controlled medications, and order labs and referrals. Tools: start_conversation, send_message, wait_for_reply, get_conversation, list_conversations, get_visit_notes, cancel_request.
• Prescriptions. Read what was prescribed and where it was sent. Tools: get_prescriptions, get_prescription.
• Health records. Link the patient's US insurance and read coverage, conditions, medications, allergies, immunizations, labs, vitals, visits, and claims. Tools: start_flexpa_link, complete_flexpa_import, get_health_records.
• Care prices. Estimate what a service costs under the patient's plan, check whether a named provider is in network, and pull public evidence about a provider. Tools: search_care_prices, check_network_status, get_provider_evidence. Also available without an account through the Pricing API.
• Care requests. Propose a referral move, an appointment booking, or a reschedule. The patient approves on Arlo's hosted sheet and Arlo's care team executes it. Tools: create_care_request, get_care_request, list_care_requests.
• Payments and webhooks. get_payment_status, start_visit_payment, and register_webhook for agents that want Arlo to wake them when a clinician replies instead of polling.

How a visit works

1. Connect. The agent connects to the server and the patient signs in once with OAuth.
2. Trigger. When the patient describes a health concern, the agent confirms their current state and calls start_conversation.
3. Triage. Arlo's AI triage asks follow-up questions. The agent relays them and answers with send_message.
4. Pay. The patient reviews the summary and approves a temporary hold for the visit on a Stripe link. No card setup step.
5. Treat. A licensed clinician reviews the case, can ask more questions, and can prescribe or order labs and referrals. The conversation stays open for follow-ups.

Connect it

• Claude: add Arlo as a custom connector.
• ChatGPT: add Arlo as an app with developer mode.
• OpenClaw: install the arlo-health plugin.
• Custom agents: follow the quickstart, or use the REST API if your runtime cannot speak MCP.

Where it works

Clinician visits are available to patients who are physically in California. The care pricing tools work anywhere in the US for supported insurers. Arlo is not for emergencies, mental health crises, controlled substances, or conditions that need a physical exam.