healthclaw-guardrails — Claude Code MCP server: setup guide
Security layer between AI agents and clinical FHIR servers. It redacts personal health identifiers, keeps audit trails, and isolates tenants.
claude mcp add --transport http healthclaw-guardrails https://mcp-server-production-5112.up.railway.app/mcpWhat healthclaw-guardrails does
This server places defensive boundaries between reasoning models and healthcare databases. It scrubs protected health information on the fly, isolates multi-tenant environments, and forces step-up authentication before sensitive actions execute. Both FHIR R4 US Core v9 and R6 specifications receive support.
Inside are 29 distinct MCP tools. The package includes $interpret for explaining lab results, $care-gaps for preventive screening alerts, and structured questionnaire extraction. The codebase targets Python 3.11+ and provides Docker Compose configurations for local hosting.
Live deployments reject unauthenticated connections. A deployment-scoped bearer token in the Authorization header is mandatory for the production URL. For preliminary evaluations, a public demo server runs synthetic test records without credential checks.
What it helps with
- strip personal health information before passing clinical records to an LLM
- query FHIR servers safely through MCP clients
- explain laboratory tests via the
$interpretoperation - track treatment gaps using
$care-gapsendpoints
Example prompts for the agent
- Search my medical records for recent lab results and explain what they mean.
- Check for missing screenings and appointments using the
$care-gapstool. - Populate the clinical questionnaire using
$populateand show the draft.
How to install healthclaw-guardrails
Platform CLI: remote server
claude mcp add --transport http healthclaw-guardrails https://mcp-server-production-5112.up.railway.app/mcpPlatform CLI: remote server
codex mcp add healthclaw-guardrails --url https://mcp-server-production-5112.up.railway.app/mcpConfig file
{
"mcpServers": {
"healthclaw-guardrails": {
"url": "https://mcp-server-production-5112.up.railway.app/mcp"
}
}
}Config file
{
"servers": {
"healthclaw-guardrails": {
"type": "http",
"url": "https://mcp-server-production-5112.up.railway.app/mcp"
}
}
}Platform CLI: remote server
gemini mcp add -t http healthclaw-guardrails https://mcp-server-production-5112.up.railway.app/mcpStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about healthclaw-guardrails
How do I install healthclaw-guardrails in Claude Code?
Run claude mcp add --transport http healthclaw-guardrails https://mcp-server-production-5112.up.railway.app/mcp in your terminal.
What is healthclaw-guardrails?
It is an MCP guardrail layer that strips sensitive health identifiers and audits access whenever AI agents interact with FHIR backends.
How does the demo URL differ from production?
The public demo endpoint serves strictly synthetic records without credentials, while the production server requires a valid Authorization: Bearer <token> header.
