Deployment¶
Deployment Options¶
keycloak-agent supports local stdio, a loopback-only development listener, a
least-privilege stdio container, and a remote authenticated HTTPS boundary.
Provider endpoint, credential, selector, identity, and trust material are supplied
at runtime through AgentConfig; none is stored in this repository.
Installed stdio process¶
{
"mcpServers": {
"keycloak": {
"command": "keycloak-mcp",
"args": [],
"env": {"MCP_TOOL_MODE": "intent"}
}
}
}
Loopback development listener¶
Do not expose this listener beyond loopback. Network deployments require direct TLS
or an explicitly trusted TLS-terminating ingress, configured authentication, exact
MCP_ALLOWED_HOSTS, and an exact trusted-proxy CIDR policy.
Least-privilege local container¶
docker run -i --rm \
--read-only \
--cap-drop=ALL \
--security-opt=no-new-privileges \
--pids-limit=256 \
--tmpfs /tmp:rw,noexec,nosuid,nodev,size=64m \
-e TRANSPORT=stdio \
registry.example.invalid/keycloak-agent@sha256:<digest> keycloak-mcp
The operator projects the selected AgentConfig profile into the process at runtime; the image remains immutable and contains no environment connection profile.
Remote authenticated HTTPS endpoint¶
Store the real remote URL, outbound identity reference, and TLS-profile reference in
AgentConfig, not in MCP client JSON or documentation.
This page covers running keycloak-agent as a long-lived service: the transports,
the optional agent server, a Docker Compose stack, putting it behind a Caddy reverse
proxy, and giving it a DNS name with Technitium. To provision the Keycloak
server it connects to, see Backing Platform.
keycloak-agentships two console scripts: an MCP server (keycloak-mcp) exposing the typed tool surface, and an optional Pydantic-AI agent server (keycloak-agent) that consumes those tools. Deploy the MCP server on its own for a policy router / agent to call, or add the agent server for autonomous operation.
Run the MCP server¶
The transport is selected with --transport (or the TRANSPORT env var):
Health check (HTTP transports):
Configuration (environment)¶
keycloak-agent is configured entirely from the environment. The required set:
| Var | Default | Meaning |
|---|---|---|
KEYCLOAK_URL |
Required | Keycloak base admin URL |
KEYCLOAK_AGENT_USERNAME |
Required for password authentication | Admin account username |
KEYCLOAK_AGENT_PASSWORD |
Required for password authentication | Admin account password |
KEYCLOAK_REALM |
master |
Realm name |
Plus HOST / PORT / TRANSPORT for HTTP transports. The full set is documented
in .env.example;
copy it to .env and fill in your values. The server registers its tools and
remains inactive on administrative writes when credentials are absent.
Docker Compose¶
The repo ships docker/compose.yml.
It reads a sibling .env and publishes the HTTP server on :8000:
services:
keycloak-agent:
image: example/keycloak-agent@sha256:<digest>
container_name: keycloak-agent
hostname: keycloak-agent
restart: always
env_file:
- .env
environment:
- PYTHONUNBUFFERED=1
- HOST=0.0.0.0
- PORT=8000
- TRANSPORT=streamable-http
ports:
- "8000:8000"
healthcheck:
test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8000/health')"]
interval: 30s
timeout: 10s
retries: 3
cp .env.example .env # then edit KEYCLOAK_* values
docker compose -f docker/compose.yml up -d
docker compose -f docker/compose.yml logs -f
Run the agent server¶
The optional agent server (keycloak-agent console script) starts a Pydantic-AI
agent that consumes the MCP tools. Point it at a running MCP server with --mcp-url,
or let it launch one from a bundled mcp_config.json:
# Connect the agent to an already-deployed MCP HTTP server
keycloak-agent --mcp-url http://keycloak-agent:8000/mcp --host 0.0.0.0 --port 9000
The agent server listens on :9000 by default. A companion Compose file deploys it
alongside the MCP server, wiring MCP_URL to the MCP service by container name:
services:
keycloak-agent-server:
image: example/keycloak-agent@sha256:<digest>
container_name: keycloak-agent-server
hostname: keycloak-agent-server
restart: always
entrypoint: ["keycloak-agent"]
env_file:
- .env
environment:
- MCP_URL=http://keycloak-agent:8000/mcp
- HOST=0.0.0.0
- PORT=9000
ports:
- "9000:9000"
depends_on:
- keycloak-agent
Behind a Caddy reverse proxy¶
Expose the HTTP server on a hostname with automatic TLS. Add to your Caddyfile:
# Internal (self-signed) — homelab .example.invalid zone
keycloak-agent.example.invalid {
tls internal
reverse_proxy keycloak-agent:8000
}
Reload Caddy:
DNS with Technitium¶
Point the hostname at the host running Caddy. Via the Technitium API:
curl -s "http://technitium.example.invalid:5380/api/zones/records/add" \
--data-urlencode "token=$TECHNITIUM_DNS_TOKEN" \
--data-urlencode "domain=keycloak-agent.example.invalid" \
--data-urlencode "zone=arpa" \
--data-urlencode "type=A" \
--data-urlencode "ipAddress=192.0.2.10" \
--data-urlencode "ttl=3600"
…or add an A record keycloak-agent.example.invalid → <caddy-host-ip> in the Technitium web
console (http://technitium.example.invalid:5380). The ecosystem
technitium-dns-mcp automates
this as a tool.
Register with an MCP client¶
Add to your client's mcp_config.json (multiplexer nickname key):
{
"mcpServers": {
"keycloak_agent": {
"command": "uv",
"args": ["run", "keycloak-mcp"],
"env": {
"KEYCLOAK_URL": "<configured-endpoint>",
"KEYCLOAK_AGENT_USERNAME": "<configured-principal>",
"KEYCLOAK_AGENT_PASSWORD": "<runtime-secret>",
"KEYCLOAK_REALM": "master"
}
}
}
}
For a remote HTTP server, point the client at http://keycloak-agent.example.invalid/mcp instead.