Connect clientsGuide

Connect other clients

Use any OpenAI-compatible tool, SDK, or script against this gateway.

By the end: Any client that accepts a custom base URL returns model replies through RouteForge.

Before you start

You have created an API key
The client lets you set a base URL

Any client that accepts a custom base URL can use RouteForge. This page gives the values, a request you can paste, and the mistakes that account for most setup failures.

Common values

Field	Value	Notes
OpenAI-compatible base URL	http://localhost:3000/v1	For editors, CLI agents, LangChain, and OpenAI SDKs.
Anthropic-compatible base URL	http://localhost:3000	For tools that take an Anthropic endpoint.
API key	sk_live_...	From the dashboard. No spaces or line breaks.
Model	Exact catalog ID	Case-sensitive. Find it on the models page.

Request

cURL

curl http://localhost:3000/v1/chat/completions \
  -H "Authorization: Bearer $ROUTEFORGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.6-sol",
    "messages": [{ "role": "user", "content": "Reply in one sentence that RouteForge is connected." }]
  }'

Python

import os
from openai import OpenAI

client = OpenAI(base_url="http://localhost:3000/v1", api_key=os.environ["ROUTEFORGE_API_KEY"])
res = client.chat.completions.create(
    model="gpt-5.6-sol",
    messages=[{"role": "user", "content": "Reply in one sentence that RouteForge is connected."}],
)
print(res.choices[0].message.content)

Common mistakes

The base URL ends up as /v1/v1/chat/completions. The value already includes /v1, so a client that appends its own path must be given the root instead.
The Authorization header is missing Bearer, or the tool expects only the raw key.
The model name came from another platform. Use an ID from this gateway's list.
A tool that only speaks one protocol is given a model from another family. Check protocol compatibility.

Client guides

Dedicated setup pages exist for Claude Code, Codex, Cursor and TRAE, OpenCode, and other tools.