Janitor AI proxy setup with Plan2API

Janitor AI proxy settings at a glance

When you set up a Janitor AI custom proxy, the configuration form asks for five fields. The table below shows each field and the exact value to enter.

Add Plan2API as a proxy in Janitor AI

1. Sign up at Plan2API with a Google account, connect a Claude, ChatGPT, Google AI or Kimi subscription, and copy your p2a_ API key.
2. Open any chat in Janitor AI.
3. Click the gear icon at the top right of the chat and open API Settings (some Janitor AI versions show a menu icon at the top right instead of a gear).
4. Select Proxy and click Add Configuration.
5. Fill in the five fields from the table above, pasting the Model Name and API Key with no extra spaces.
6. Click Add, then click Save Settings.
7. Refresh the whole page (F5 on a computer, or pull down on a phone), because Janitor AI only picks up a new proxy after a page refresh.
8. Send a short test message in the chat.

Once active, the proxy replaces Janitor AI's free default JanitorLLM (JLLM) model and sends your chats to the external model instead.

Pick a model name

The prefix in the Model Name tells the API which connected subscription to route your chat through:

• claude/claude-sonnet-5: needs a Claude Pro or Max subscription and has a 1M-token context window with up to 128K output tokens.
• claude/claude-opus-5-5: needs a Claude Pro or Max subscription and gives you a 1M-token context window with up to 128K output tokens (the anthropic/ prefix and dotted names like claude-opus-5.5 work too).
• gemini/gemini-3.8-flash-high: runs through a Google AI Pro or Ultra subscription and has a 1,048,576-token context window.
• kimi/k3: needs a Kimi Code membership and has a 1,048,576-token context window (it starts on the 256K window and moves to the 1M window when a conversation passes about 192K tokens).
• openai/gpt-6.1-sol: uses a ChatGPT plan through Codex, so pick a Codex model your ChatGPT plan includes (codex/<model> works too).

Janitor AI's guides say to copy-paste the Model Name exactly, so paste it without typos, or query GET /v1/models to list every valid id your key can use.

Test your key outside Janitor AI

Run the curl request below in a terminal with your p2a_ key stored in the PLAN2API_KEY environment variable and the exact same model id you typed into Janitor AI. If the command returns a reply, your key, model id and connected subscription all work, and any remaining issue is in your Janitor AI configuration. Getting an HTTP 401 invalid_api_key error here means the key itself is missing or wrong.

curl https://api.plan2api.com/v1/chat/completions \
  -H "Authorization: Bearer $PLAN2API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude/claude-sonnet-5",
    "messages": [{"role": "user", "content": "Say hi in five words."}]
  }'

What works and what doesn't

• Janitor AI sends proxy requests directly from your browser tab, and they work without an extra proxy because Plan2API sends Access-Control-Allow-Origin: * CORS headers.
• Janitor AI's Temperature setting only takes effect on Gemini models, because Claude and Codex models ignore temperature and top_p while Kimi fixes temperature at 1 and top_p at 0.95.
• Every reply you generate counts as one request, so the Free plan's 50 chat requests a day go quickly during a long chat.
• Each provider's own content filters still apply, and an empty reply (pgshag2) or a reply cut off mid-sentence usually means a content filter, most often on Gemini models.
• One p2a_ key can hold several connected subscriptions across different providers, so switching from Claude to Gemini or Kimi only takes changing the Model Name field.
• Plan2API adds no token quota of its own, though your underlying subscription's usage limits still apply.

Fix Janitor AI proxy errors

Before troubleshooting specific error codes, run through Janitor AI's Fix-It First routine:

1. Open your proxy configuration and check that it says Active (if it doesn't, click it to activate).
2. Click Save Settings even if you didn't change anything.
3. Refresh the entire Janitor AI page.
4. Re-open the configuration and click Test.

If the chat still fails after a reload, match what you see on screen to the symptom and fix in the table below.

Use your Claude or ChatGPT plan in Janitor AI

Plan2API turns a Claude Pro or Max, ChatGPT, Google AI Pro or Ultra, or Kimi subscription you already pay for into one OpenAI-compatible p2a_ key. You keep paying your provider for your subscription, and Plan2API never bills per token.

Your subscription's own usage limits apply alongside your plan's request limits. If you connect more than one account from the same provider, a request that hits a limit moves to the next connected account before any output is sent.

• Free: $0, 1 connected subscription, 50 chat requests a day (resets at 00:00 UTC), 1 request per second with bursts up to 20, 4 requests at once, no card needed.
• Pro: $12 a month or $99 a year (about $8.25 a month, 31% less than monthly), 3 connected subscriptions, no daily request cap, 10 requests per second with bursts up to 600, 16 requests at once.
• Scale: $29 a month or $249 a year (about $20.75 a month, 28% less), 25 connected subscriptions, no daily request cap, 10 requests per second with bursts up to 600, 32 requests at once; see pricing for full details.