Umbryn MCP runs on your own machine — the client starts it, so there is no endpoint to ping. 120 installs a week from pypi. Last commit 14 Aug 2026.
Redacts PII/PHI from text before it reaches an LLM. Fail-closed, self-hosted, no egress.
Today is the operative word: we check Umbryn MCP every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add umbryn-mcp -- uvx umbryn-mcp
{
"mcpServers": {
"umbryn-mcp": {
"args": [
"umbryn-mcp"
],
"command": "uvx"
}
}
}
[mcp_servers.umbryn-mcp]
command = "uvx"
args = ["umbryn-mcp"]
{
"mcpServers": {
"umbryn-mcp": {
"args": [
"umbryn-mcp"
],
"command": "uvx"
}
}
}
{
"mcpServers": {
"umbryn-mcp": {
"args": [
"umbryn-mcp"
],
"command": "uvx"
}
}
}
This one needs environment variables set before it will start:
UMBRYN_ENGINE (Detection engine: auto (default), regex, or presidio), UMBRYN_MIN_CONFIDENCE (Trust threshold (0-1); detections below it fail closed. Default 0.5), UMBRYN_DETECTION_FLOOR (Sensitivity floor (0-1); below this a signal is noise. Default 0.35).
The author declared them in the registry entry; get the values from the project itself.
Redact PII from text before it reaches a model. Nothing stored, no third-party AI.
Detect prompt injection, jailbreaks, and code injection in untrusted text before it reaches an LLM.
Prompt-injection detector for untrusted text before it reaches an LLM (RU/EN). BYO lovec.tech key.
Detect and redact PII, PHI, PCI and secrets in text and files before an AI agent sees them.
Detects and redacts PII (emails, phones, SSNs, names, addresses) from text. $0.02/call via x402.
Read-only, self-hosted MCP access to IMAP email, mirrored locally and indexed by notmuch.
Permanently redact PII from PDFs and scans. EU/Swiss-hosted, never used to train AI models.
Prompt-injection firewall for AI agents — scan untrusted text before LLM calls.
Answers built from our own checks of this server.