Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
/**
* Integration: MCP Proxy routing
*
* Validates centralized HTTP routing through mcp-proxy to MCP servers.
* Requires: minikube cluster with mcp-proxy exposed (via port-forward or ingress).
*
* Note: MCP_PROXY_ENABLED=true must be set on mcp-host for proxy routing.
*/
import { describe, it, expect, beforeAll } from "vitest";
import { MCP_PROXY_URL, isServiceUp, fetchJson, postJson, bearer } from "./helpers.integration.js";
let proxyUp = false;
beforeAll(async () => {
proxyUp = await isServiceUp(MCP_PROXY_URL);
if (!proxyUp) {
console.log("[mcp-proxy-routing] mcp-proxy not available — tests will be skipped");
}
});
describe("MCP Proxy — health", () => {
it("returns 200 on /health", async () => {
if (!proxyUp) return;
const { status } = await fetchJson(`${MCP_PROXY_URL}/health`);
expect(status).toBe(200);
});
});
describe("MCP Proxy — server list", () => {
it("GET /servers returns 401 without auth token", async () => {
if (!proxyUp) return;
const { status } = await fetchJson(`${MCP_PROXY_URL}/servers`);
expect([401, 403]).toContain(status);
});
it("GET /servers with valid token returns array (may be empty)", async () => {
const testToken = process.env.TEST_MCP_PROXY_TOKEN;
if (!proxyUp || !testToken) return;
const { status, data } = await fetchJson<unknown[]>(
`${MCP_PROXY_URL}/servers`,
{ headers: bearer(testToken) }
);
expect(status).toBe(200);
expect(Array.isArray(data)).toBe(true);
});
});
describe("MCP Proxy — server routing", () => {
it("POST to non-existent server returns 404", async () => {
if (!proxyUp) return;
const testToken = process.env.TEST_MCP_PROXY_TOKEN;
if (!testToken) return;
const { status } = await postJson(
`${MCP_PROXY_URL}/servers/non-existent-server/mcp`,
{ jsonrpc: "2.0", id: "1", method: "tools/list" },
bearer(testToken)
);
expect([404, 503]).toContain(status);
});
});