paperless-mcp: deploy MCP for AI document search

- New stack `paperless-mcp` running barryw/PaperlessMCP v0.1.19 (.NET,
  HTTP+SSE on :5000) wraps paperless-ngx's built-in FTS. 43 tools exposed.
- In-cluster only egress to paperless-ngx svc; no Cloudflare hop on
  MCP-internal traffic.
- Read-only at paperless layer: dedicated `claude-mcp` user (non-superuser)
  in new `claude-mcp-readers` group with view-only Django perms; existing
  279 docs bulk-granted view perm via /api/documents/bulk_edit/;
  workflow #2 auto-grants the group on new docs (Consumption Added).
- Gateway-level bearer auth via new Traefik plugin
  Aetherinox/traefik-api-token-middleware@v0.1.4 (loaded in traefik stack
  alongside crowdsec-bouncer); per-stack Middleware CRD `bearer-auth`
  pulls token list from Vault `secret/paperless-mcp/bearer_tokens`.
- Vault `secret/paperless-mcp` holds: paperless_api_token (synced to
  K8s Secret via ESO; pod env via secret_key_ref), bearer_tokens
  (JSON array, read at plan time), bearer_token_viktor_laptop (mirror
  for laptop wiring), paperless_user_password (paperless UI fallback).
- Image auto-update via Keel (semver minor policy, hourly poll).
- Ingress dns_type=proxied → Uptime Kuma external monitor auto-created
  by external-monitor-sync CronJob.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
Viktor Barzin 2026-05-17 11:14:35 +00:00 committed by Viktor Barzin
parent 2eb611fd6d
commit 9521bb0b17
5 changed files with 267 additions and 1 deletions

View file

@ -53,6 +53,7 @@
| insta2spotify | Instagram reel song ID to Spotify playlist | insta2spotify |
| trading-bot | Event-driven trading with sentiment analysis | trading-bot |
| claude-memory | Persistent memory MCP server | claude-memory |
| paperless-mcp | Paperless-ngx document search MCP (barryw/PaperlessMCP). Traefik bearer auth via Aetherinox api-token-middleware. `auth=none` at ingress; gateway-level bearer enforced by `paperless-mcp/bearer-auth` Middleware CRD. Tokens + paperless API token in Vault `secret/paperless-mcp`. | paperless-mcp |
| council-complaints | Islington civic reporting pilot | council-complaints |
## Optional

View file

@ -0,0 +1,225 @@
variable "tls_secret_name" {
type = string
sensitive = true
}
# Vault read: bearer_tokens (JSON array, used directly in the Middleware CRD)
# and paperless_api_token (synced to a K8s Secret by ESO and consumed by the
# pod as an env var).
data "vault_kv_secret_v2" "secrets" {
mount = "secret"
name = "paperless-mcp"
}
resource "kubernetes_namespace" "paperless-mcp" {
metadata {
name = "paperless-mcp"
labels = {
tier = local.tiers.aux
"keel.sh/enrolled" = "true"
}
}
lifecycle {
# KYVERNO_LIFECYCLE_V1: goldilocks-vpa-auto-mode ClusterPolicy stamps this label on every namespace
ignore_changes = [metadata[0].labels["goldilocks.fairwinds.com/vpa-update-mode"]]
}
}
# Paperless API token (MCP -> paperless). Synced from Vault to a K8s Secret
# by ESO; the pod reads it via secret_key_ref.
resource "kubernetes_manifest" "external_secret" {
manifest = {
apiVersion = "external-secrets.io/v1beta1"
kind = "ExternalSecret"
metadata = {
name = "paperless-mcp-secrets"
namespace = "paperless-mcp"
}
spec = {
refreshInterval = "15m"
secretStoreRef = {
name = "vault-kv"
kind = "ClusterSecretStore"
}
target = {
name = "paperless-mcp-secrets"
}
data = [{
secretKey = "paperless_api_token"
remoteRef = {
key = "paperless-mcp"
property = "paperless_api_token"
}
}]
}
}
depends_on = [kubernetes_namespace.paperless-mcp]
}
module "tls_secret" {
source = "../../modules/kubernetes/setup_tls_secret"
namespace = kubernetes_namespace.paperless-mcp.metadata[0].name
tls_secret_name = var.tls_secret_name
}
# Gateway-level bearer auth. barryw/PaperlessMCP has no native auth; this
# Middleware enforces Authorization: Bearer <token> at Traefik before any
# request reaches the pod. Token list lives in Vault as a JSON array string;
# rotation = update Vault then re-apply this stack.
resource "kubernetes_manifest" "bearer_middleware" {
manifest = {
apiVersion = "traefik.io/v1alpha1"
kind = "Middleware"
metadata = {
name = "bearer-auth"
namespace = kubernetes_namespace.paperless-mcp.metadata[0].name
}
spec = {
plugin = {
# Inner key must match the static-config key in Traefik
# experimental.plugins.api-token-middleware.
api-token-middleware = {
authenticationHeader = false
bearerHeader = true
bearerHeaderName = "Authorization"
tokens = jsondecode(data.vault_kv_secret_v2.secrets.data["bearer_tokens"])
removeHeadersOnSuccess = true
authenticationErrorMsg = "Access Denied"
}
}
}
}
}
resource "kubernetes_deployment" "paperless-mcp" {
metadata {
name = "paperless-mcp"
namespace = kubernetes_namespace.paperless-mcp.metadata[0].name
labels = {
app = "paperless-mcp"
tier = local.tiers.aux
}
annotations = {
"reloader.stakater.com/auto" = "true"
"keel.sh/policy" = "minor"
"keel.sh/trigger" = "poll"
"keel.sh/pollSchedule" = "@every 1h"
}
}
spec {
replicas = 1
selector {
match_labels = {
app = "paperless-mcp"
}
}
template {
metadata {
labels = {
app = "paperless-mcp"
}
}
spec {
container {
name = "paperless-mcp"
image = "ghcr.io/barryw/paperlessmcp:v0.1.19"
port {
container_port = 5000
}
env {
name = "PAPERLESS_BASE_URL"
value = "http://paperless-ngx.paperless-ngx.svc.cluster.local"
}
env {
name = "PAPERLESS_API_TOKEN"
value_from {
secret_key_ref {
name = "paperless-mcp-secrets"
key = "paperless_api_token"
}
}
}
env {
name = "MCP_PORT"
value = "5000"
}
# barryw exposes no HTTP /health; the ping/capabilities probes are
# MCP JSON-RPC over /mcp. TCP-socket probe is what the upstream
# k8s/deployment.yaml uses.
startup_probe {
tcp_socket {
port = 5000
}
failure_threshold = 30
period_seconds = 2
}
liveness_probe {
tcp_socket {
port = 5000
}
initial_delay_seconds = 5
period_seconds = 30
}
readiness_probe {
tcp_socket {
port = 5000
}
initial_delay_seconds = 3
period_seconds = 10
}
resources {
requests = {
memory = "256Mi"
cpu = "10m"
}
limits = {
memory = "256Mi"
}
}
}
}
}
}
lifecycle {
ignore_changes = [
spec[0].template[0].spec[0].container[0].image, # Keel-managed
spec[0].template[0].spec[0].dns_config, # KYVERNO_LIFECYCLE_V1: Kyverno admission webhook mutates dns_config with ndots=2
metadata[0].annotations["keel.sh/policy"],
metadata[0].annotations["keel.sh/trigger"],
metadata[0].annotations["keel.sh/pollSchedule"],
]
}
}
resource "kubernetes_service" "paperless-mcp" {
metadata {
name = "paperless-mcp"
namespace = kubernetes_namespace.paperless-mcp.metadata[0].name
labels = {
app = "paperless-mcp"
}
}
spec {
selector = {
app = "paperless-mcp"
}
port {
name = "http"
port = 80
target_port = 5000
}
}
}
module "ingress" {
source = "../../modules/kubernetes/ingress_factory"
# auth = "none": barryw/PaperlessMCP has no native auth; the bearer-auth
# Middleware CRD attached below enforces Authorization: Bearer at Traefik.
auth = "none"
dns_type = "proxied"
namespace = kubernetes_namespace.paperless-mcp.metadata[0].name
name = "paperless-mcp"
tls_secret_name = var.tls_secret_name
homepage_enabled = false
extra_middlewares = ["paperless-mcp-bearer-auth@kubernetescrd"]
}

View file

@ -0,0 +1 @@
../../secrets

View file

@ -0,0 +1,27 @@
include "root" {
path = find_in_parent_folders()
}
dependency "platform" {
config_path = "../platform"
skip_outputs = true
}
dependency "vault" {
config_path = "../vault"
skip_outputs = true
}
# Aetherinox bearer middleware must be loaded in Traefik before our
# Middleware CRD can be applied with a non-zero token list.
dependency "traefik" {
config_path = "../traefik"
skip_outputs = true
}
# We point PAPERLESS_BASE_URL at the in-cluster service to avoid the
# Cloudflare->Traefik hop on every MCP call.
dependency "paperless-ngx" {
config_path = "../paperless-ngx"
skip_outputs = true
}

View file

@ -70,7 +70,10 @@ resource "helm_release" "traefik" {
"mkdir -p \"$STORAGE/archives/github.com/maxlerebourg/crowdsec-bouncer-traefik-plugin\"; ",
"wget -q -T 30 -O \"$STORAGE/archives/github.com/maxlerebourg/crowdsec-bouncer-traefik-plugin/v1.4.2.zip\" ",
"\"https://github.com/maxlerebourg/crowdsec-bouncer-traefik-plugin/archive/refs/tags/v1.4.2.zip\"; ",
"printf '{\"github.com/maxlerebourg/crowdsec-bouncer-traefik-plugin\":\"v1.4.2\"}' ",
"mkdir -p \"$STORAGE/archives/github.com/Aetherinox/traefik-api-token-middleware\"; ",
"wget -q -T 30 -O \"$STORAGE/archives/github.com/Aetherinox/traefik-api-token-middleware/v0.1.4.zip\" ",
"\"https://github.com/Aetherinox/traefik-api-token-middleware/archive/refs/tags/v0.1.4.zip\"; ",
"printf '{\"github.com/maxlerebourg/crowdsec-bouncer-traefik-plugin\":\"v1.4.2\",\"github.com/Aetherinox/traefik-api-token-middleware\":\"v0.1.4\"}' ",
"> \"$STORAGE/archives/state.json\"; ",
"echo \"Plugins pre-downloaded successfully\"",
])]
@ -176,6 +179,15 @@ resource "helm_release" "traefik" {
moduleName = "github.com/maxlerebourg/crowdsec-bouncer-traefik-plugin"
version = "v1.4.2"
}
# Static-token bearer/header auth middleware. Used by services that
# need gateway-level API-key/bearer enforcement without app-layer auth
# (e.g. paperless-mcp, which has no native auth). Plugin key
# `api-token-middleware` is the name to use as the inner key in
# `Middleware.spec.plugin.<key>` on consuming Middleware CRDs.
api-token-middleware = {
moduleName = "github.com/Aetherinox/traefik-api-token-middleware"
version = "v0.1.4"
}
}
}