LobeChat

LobeChat is a popular open-source web AI client with both hosted and self-hosted options. Point its OpenAI provider at Tokensmart.

💡 Quick setup: create a key on API Keys; the API Proxy URL value below is ready to paste.

Hosted

1. Open chat.lobehub.com
2. Avatar (top right) → Settings
3. Left pane → Language Model → OpenAI
4. Fill in:

5. Click Check
6. In Model List below, enable or add gpt-5.4 / claude-sonnet-4-6 etc.

Self-hosted (Docker)

docker run -d \
  -p 3210:3210 \
  -e OPENAI_API_KEY=pk_live_xxxxxxxxxxxxxxxx \
  -e OPENAI_PROXY_URL=https://api.tokensmart.ai/v1 \
  --name lobe-chat \
  lobehub/lobe-chat

Tokensmart tips

• Team sharing pattern A — one key, many users: let LobeChat's own user accounts handle auth; use a single Tokensmart key behind the scenes. You only manage one balance and one billing stream.
• Team sharing pattern B — one key per user: if you want per-user accounting, hand each teammate/family member their own Tokensmart key with a personal daily limit, and deploy a multi-user LobeChat so they can enter their own.
• Add new models manually: LobeChat's built-in model list updates slower than Tokensmart ships new models (typically 24–48h after release). When something new lands, paste the Tokensmart model ID under Model List → Custom Model.
• Self-host with env vars: passing OPENAI_PROXY_URL at container launch is more robust than the UI — upgrading the Docker image will never clobber your config.

Troubleshooting

• CORS in browser console — the hosted version proxies through its own backend and should not hit CORS; if you self-host without a backend, use the Docker build
• Only some models showing — LobeChat only lists models it has internally curated. Add missing ones under Model List → Custom Model