Claude, GPT, Gemini and Kimi in Xcode chat with Plan2API

Xcode chat provider settings at a glance

When you add Plan2API as an Internet Hosted provider, the dialog asks for four values. The table below shows what to enter in each field.

Add Plan2API to Xcode

1. Sign up at Plan2API with a Google account, connect a subscription, and copy your p2a_ key.
2. Make sure you're running Xcode 26.3 or later, because 26.3 fixed a bug where custom model providers disappeared between launches of Xcode.
3. Choose Xcode > Settings from the menu bar and select Intelligence in the sidebar.
4. Under Chat, click Add a Chat Provider (some Xcode 26 releases label this button Add a Model Provider) and select Internet Hosted.
5. Enter the URL, API Key with the Bearer prefix and a space before your key, API Key Header, and Description from the table above, then click Add.
6. Select Plan2API under Chat so Xcode fetches the model list, turn on the models you want to use, and star the ones you use most so they stay at the top of the menu.
7. Open a project, click the Coding Assistant button or press Command-0, click New Conversation, and choose a Plan2API model under Chat.
8. Send a short prompt to test the setup; the placeholder text in the message field names the model the conversation is using.

Pick a model for the coding assistant

Xcode's coding assistant only works with models that support tool calling and ideally have 200K tokens of context or more, and every Plan2API model supports tool calling.

• claude/claude-sonnet-5 uses a Claude Pro or Max subscription and has a 1M-token context window with up to 128K output.
• claude/claude-opus-5-5 uses a Claude Pro or Max subscription and has a 1M-token context window with up to 128K output.
• openai/gpt-6.1-sol uses a ChatGPT plan through Codex, so pick a Codex model your plan includes.
• gemini/gemini-3.8-flash-high uses a Google AI Pro or Ultra subscription and has a 1,048,576-token context window.
• kimi/k3 uses a Kimi Code membership and has a 1,048,576-token context window, starting on the 256K window and moving to 1M when a conversation passes about 192K tokens.

Xcode lists every id returned by GET /v1/models, so star a few favorites in the provider settings to keep the model menu short.

Check the key outside Xcode

Xcode calls two endpoints on a chat provider: GET /v1/models for the model list and POST /v1/chat/completions for messages. The curl commands below send those exact requests using your key from the PLAN2API_KEY environment variable. If both requests return valid responses in your terminal, your key and connected subscription work, and any issue is in the Xcode dialog settings; if you get a 401 invalid_api_key error here, the key itself is wrong or missing.

# The two endpoints Xcode calls, with the same key and header
curl https://api.plan2api.com/v1/models \
  -H "Authorization: Bearer $PLAN2API_KEY"

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 ready"}]}'

What works and what doesn't

• Chat conversations in the coding assistant work, and so do the source editor's coding tools (Explain, Document, Generate a Playground, Generate a Preview, and Generate Fix for Issue), which all send requests to your current chat model.
• Tool calling and streaming work across Claude, Codex, Gemini, and Kimi models.
• Xcode's agents (Claude Agent, Codex, Gemini, and agents added via the Agent Client Protocol with Add an Agent) don't use chat providers and keep their own separate sign-in or configuration.
• Xcode has no context-window setting for chat providers and keeps Project Context on by default, so turn Project Context off in the message field and reference specific files and symbols with @ when you want to send smaller requests.
• Each turn re-sends the whole conversation and project context, so start a new conversation when you move to a new task.
• On Plan2API's Free plan, every chat prompt and every coding-tool action in the source editor counts toward the limit of 50 chat requests a day.

Fix Xcode chat provider errors

When the Plan2API provider fails to load models or answer prompts, the cause is usually a value in the dialog or an account limit. The table below maps each symptom and error code to its fix.

Use your Claude or ChatGPT plan in Xcode

Plan2API turns a Claude Pro or Max, ChatGPT (through Codex), Google AI Pro or Ultra, or Kimi subscription you already pay for into one p2a_ API key, so one chat provider in Xcode can serve all four model families once each subscription is connected.

You keep paying your provider for your subscription, Plan2API never bills per token, and your subscription's own usage limits apply. If you connect several accounts from the same provider, a request that hits a limit on one account 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), no card required.
• Pro: $12 a month or $99 a year, 3 connected subscriptions, no daily request cap.
• Scale: $29 a month or $249 a year, 25 connected subscriptions, no daily request cap; see pricing for full plan details.