OpenRouter

Lightweight, dependency-free, in-memory OpenRouter API fake. OpenRouter is OpenAI-compatible, so this works with the official openai SDK pointed at it. All output is deterministic (hash-derived) and SSE streaming is supported.

Default port: 4861

Quick start

import { OpenrouterServer } from "./services/openrouter/src/server.js";

const server = new OpenrouterServer(4861);
await server.start();
// ... run your app/tests ...
await server.stop();

Point the openai SDK at it:

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "sk-or-parlel",
  baseURL: "http://127.0.0.1:4861/api/v1",
});

const res = await client.chat.completions.create({
  model: "anthropic/claude-3.5-sonnet",
  messages: [{ role: "user", content: "hello" }],
});
// res.choices[0].message.content => deterministic text; res.provider => routing field

Access via MCP / preview URL

Base URL: http://127.0.0.1:4861/api/v1
Health: GET /health → { "status": "ok" }
Auth: Authorization: Bearer <key> (any non-empty token). GET /api/v1/models is public.

Implemented operations

POST /api/v1/chat/completions — OpenAI-compatible chat. Supports stream: true (SSE, terminated by data: [DONE]). Adds an OpenRouter provider routing field.
POST /api/v1/embeddings — OpenAI-compatible embeddings (deterministic).
GET /api/v1/models — public model catalog (vendor-prefixed ids like openai/gpt-4o).

Service & inspection operations (parlel extensions)

GET / / GET /health / POST /__parlel/reset / GET /__parlel/requests.

Surface coverage

This emulator faithfully replicates the API surface most application code and agents exercise. Anything below the supported lines is either an intentional design choice for a fast, zero-cost local emulator (✓ By design) or a candidate for a future release (⟳ Roadmap) — never a silent inaccuracy.

Legend: ✅ fully supported · ◐ accepted (stored, not strictly enforced) · ✓ by design · ⟳ on the roadmap.

Feature	Status
`chat.completions` (non-stream + SSE stream)	✅ Supported
`embeddings`	✅ Supported
`models` list with routing/provider field	✅ Supported
Deterministic, reproducible output	✅ Supported
Real model inference	✓ By design — Deterministic stub output — repeatable assertions, no API spend
Provider preferences / fallbacks / transforms	◐ `provider` echoed from model prefix only
Credits / cost accounting	✓ By design — Not enforced
Tool/function calling, vision	◐ Accepted, not specially handled

Error codes & shapes

Errors use the OpenAI envelope: { "error": { "message", "type", "code" } }.

Status	When
`401`	missing/invalid `Authorization` (except `/models`)
`400`	missing `model`/`messages`/`input` or bad JSON
`404`	unknown endpoint

Manifest

See services/openrouter/manifest.json:

name: openrouter, port: 4861, protocol: http, healthcheck: /health, startup ≈ 100ms
env: OPENROUTER_API_KEY, OPENROUTER_BASE_URL

Configuration — `test.env`

Copy these into your test.env (used by the bridge sidecar flow). Tokens are Parlel's seeded test credentials — any non-empty value is accepted by the emulator, so you rarely need to change them. Swap in real credentials only when pointing at the live service in prod.env.

OPENROUTER_API_KEY=sk-or-parlel
OPENROUTER_BASE_URL=http://parlel-bridge:4861/api/v1