LobsterAI custom model: your own API key for Claude, GPT and Chinese models (November 2026)
Can LobsterAI use your own API key? Yes. As of 2026-11-10, per the official README, the 2026.11.6 source and a maintainer's reply: open Settings → Custom Model, click + Add Custom at the bottom of the provider list, choose Anthropic Compatible (Messages) or OpenAI Compatible (Chat Completions) under API Format, enter the QCode Base URL and a key that starts with cr_, then click Add Model and Test Connection. Claude uses Anthropic Compatible with https://api.qcode.cc/api; GPT uses OpenAI Compatible with https://api.qcode.cc/openai/v1; GLM, Kimi, DeepSeek and Qwen work on either. Connection successful means the address and the key are both right.
Updated 2026-11-10
Point this tool at the QCode endpoint — one key for Claude, GPT and Chinese models.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
Four things to know before you fill it in
Claude: https://api.qcode.cc/api
Set API Format to Anthropic Compatible (Messages) and API Base URL to https://api.qcode.cc/api. Test Connection appends /v1/messages itself, so enter the address as it is, without /v1 or /v1/messages. Claude models are served only on this path, and GLM, Kimi, DeepSeek and Qwen can use it too, for example claude-sonnet-5-5, claude-opus-5-5 and glm-5.3.
GPT: https://api.qcode.cc/openai/v1
Set API Format to OpenAI Compatible (Chat Completions) and API Base URL to https://api.qcode.cc/openai/v1. The address already ends in /v1, so Test Connection only appends /chat/completions. GPT models use this path, for example gpt-6.1-sol and gpt-6.1-terra; GLM, Kimi, DeepSeek and Qwen can use it too, for example deepseek-v4.1-flash, kimi-k3 and qwen3.8-max.
No /v1 for Anthropic, /v1 for OpenAI
The hint under API Format reads: Must match the protocol behind your Base URL; OpenAI-compatible URLs usually end with /v1. Under Examples (SiliconFlow, click to fill), OpenAI Compatible fills https://api.siliconflow.cn/v1 and Anthropic Compatible fills https://api.siliconflow.cn. QCode follows the same pattern: https://api.qcode.cc/openai/v1 for OpenAI and https://api.qcode.cc/api for Anthropic.
An option since 2026.10.23
The official v2026.10.23 changelog says custom models support OpenAI Responses and Gemini native APIs. With OpenAI Compatible (Responses), QCode also takes https://api.qcode.cc/openai/v1; the request lands on /openai/v1/responses and serves GPT models.
What LobsterAI is
LobsterAI (Chinese name: Youdao Lobster) is an open-source desktop AI agent from NetEase Youdao, built as an all-scenario office assistant, and it takes your own API key: as of 2026-11-10 the official README says to add your own API key under Settings → Custom Model. The official FAQ says it is built on the open-source agent framework OpenClaw, ships a one-click desktop client, and is released under the MIT License on GitHub (netease-youdao/LobsterAI); the README says OpenClaw is the agent runtime underneath and LobsterAI ships a pinned build of it. Installers are on the download page https://lobsterai.youdao.com/#/download-list and on GitHub Releases, for macOS (Apple Silicon), macOS (Intel) and Windows x64. The official sources name the menu in two ways: the FAQ and the user manual say Settings → Models, while in the 2026.11.6 source that settings tab is labelled Custom Model, with the built-in providers and + Add Custom in one list. Version: the release GitHub marks Latest is LobsterAI 2026.11.6 (published 2026-11-06), and the official home page shows 2026.10.30 as the latest version on the same day.
Since 2026.10.23: OpenAI Responses on custom models
The official changelog entry v2026.10.23 (2026-10-23) says custom models support OpenAI Responses and Gemini native APIs. The matching API Format option is OpenAI Compatible (Responses), with the hint: Requires the provider to support the OpenAI Responses API. Requests go to Base URL + /responses. On QCode, Responses serves GPT models: set Base URL to https://api.qcode.cc/openai/v1 and the request lands on /openai/v1/responses, which the QCode endpoints page lists as equivalent to /openai/responses. This setup follows the official wording and the endpoints page, and the Test Connection result is the final word; for everyday chat, OpenAI Compatible (Chat Completions) is enough. The Gemini (generateContent) option added in the same release is for Gemini's native API; for QCode, the three formats above cover everything.
Three releases that matter for custom models
The v2026.8.14 changelog raised the cap on custom model providers from 10 to 30. For QCode, one Anthropic entry and one OpenAI entry with the same cr_ key are enough.
The v2026.10.23 changelog: custom models support OpenAI Responses and Gemini native APIs. On QCode, Responses also takes https://api.qcode.cc/openai/v1.
GitHub published LobsterAI 2026.11.6. As of 2026-11-10 it is the release marked Latest, while the official home page shows 2026.10.30 on the same day; the UI labels on this page were checked against the 2026.11.6 source.
What the docs say, and what the source shows
Stated in the official docs and UI
1. Connection successful after Test Connection means the address and the key both arrived; in issue #362 (2026-10-06) the maintainer says the provider is then enabled automatically and can be picked in the model selector of the input box. 2. The official model configuration guide lists four messages: API Key Invalid, Insufficient Balance, Network Error and Service Unavailable; on QCode, copy the full cr_ key again, check your QCode balance, check the network, then check the service status. 3. The official manual states that the API key is stored only on your local device and is not uploaded to any server. 4. A Claude model under an OpenAI format gets model_not_available_on_endpoint from QCode: switch that provider's API Format to Anthropic Compatible (Messages) and the address to https://api.qcode.cc/api. 5. In issue #87 another service did not know the model ID it was sent and answered 400 (invalid param model); with QCode, enter model IDs exactly as /models lists them, for example claude-sonnet-5-5 and gpt-6.1-sol.
Not in the docs, but in the source
1. How Test Connection builds the path (2026.11.6 source): the Anthropic format sends to Base URL + /v1/messages, or only /messages when the address already ends in /v1; OpenAI Chat Completions appends /chat/completions when the address ends in a version such as /v1, and adds /v1 first otherwise; Responses appends /responses, also adding /v1 first when it is missing. 2. The built-in provider list follows the UI language: the Chinese UI shows the built-in providers tagged china, and the English UI shows another list; custom providers appear in both, so the QCode steps are the same in either language. 3. Issue #185 (opened 2026-02-28) asked for OpenAI's newer API format on custom providers, and the maintainer's answer then was to switch the UI to English and use the built-in OpenAI provider; the issue is still open as of 2026-11-10, and current status is that custom models support OpenAI Responses since 2026.10.23.
How it relates to OpenClaw and AutoClaw
LobsterAI and OpenClaw
OpenClaw is the open-source agent runtime framework; on its own it is configured in ~/.openclaw/openclaw.json or with the openclaw onboard wizard. LobsterAI is the desktop client built on it: the README says it ships a pinned OpenClaw build and renders its config from app state, so in LobsterAI you connect QCode only under Custom Model in the UI. Connecting the framework itself to QCode is covered at https://docs.qcode.cc/en/docs/ide/openclaw, the ecosystem and skill registry at /claw-ecosystem-guide, and a Gateway on a cloud server at /openclaw-on-cloud-server.
LobsterAI and AutoClaw
AutoClaw is Zhipu's desktop AI agent; its official docs call it the first one-click local OpenClaw app. Its custom models are added under Settings → Models & API → add a custom model, with the API protocol set to OpenAI, and the official blog says it supports model services compatible with the OpenAI Chat Completions protocol; on QCode that protocol serves GPT and Chinese models at https://api.qcode.cc/openai/v1. LobsterAI's custom models also offer Anthropic Compatible (Messages), which is the one Claude models use. AutoClaw setup is at https://docs.qcode.cc/en/docs/ide/autoclaw.
Connect QCode in LobsterAI in four steps
1. Install LobsterAI from the official download page or GitHub Releases, and copy a key that starts with cr_ from the QCode dashboard. 2. Open Settings → Custom Model and click + Add Custom at the bottom of the provider list: Display Name can be anything (for example QCode), API Format is Anthropic Compatible (Messages), API Base URL is https://api.qcode.cc/api, and API Key is your cr_ key. For GPT, add a second one: API Format OpenAI Compatible (Chat Completions), API Base URL https://api.qcode.cc/openai/v1, and the same key. 3. In the provider click Add Model; Model Name is any label, and Model ID must match /models character for character. On the Anthropic one you can enter claude-sonnet-5-5, claude-opus-5-5, claude-haiku-5-5, glm-5.3, deepseek-v4.1-flash, kimi-k3 and qwen3.8-max; on the OpenAI one gpt-6.1-sol, gpt-6.1-terra, glm-5.3, deepseek-v4.1-flash, kimi-k3 and qwen3.8-max. 4. Click Test Connection; after Connection successful the provider is enabled automatically, so pick the new model in the model selector of the input box and send ping. On an unstable network, swap the host for asia.qcode.cc or us.qcode.cc and keep the path. The same steps are in the QCode docs at https://docs.qcode.cc/en/docs/ide/lobsterai.
One QCode key, both compatible interfaces
As of 2026-11-10, one QCode API key gives you both Anthropic- and OpenAI-compatible endpoints: Claude models on the Anthropic endpoint https://api.qcode.cc/api, GPT models on the OpenAI endpoint https://api.qcode.cc/openai/v1 (the same address for Chat Completions and Responses), and the GLM, Kimi, DeepSeek and Qwen families on either. Usage is billed per token with per-model rates on /models. Three steps: ① register and top up (Alipay and other major payment methods are supported); ② buy a Personal Plan or a Pay-as-you-go Pack and copy your API key, which starts with cr_; ③ add the custom providers in LobsterAI as above. The same key also works in Claude Code and other tools. Not ready for a monthly plan? Pay-as-you-go Packs (/pay-as-you-go) start at $4 for a one-time quota you use within 7 days; if you already have a Personal Plan, add a booster pack when you need more quota.
Frequently asked questions
Where do I set up a custom model in LobsterAI?
Under Settings → Custom Model: click + Add Custom at the bottom of the provider list. That is how the README and the 2026.11.6 source put it; the official FAQ and the user manual say Settings → Models, and in the 2026.11.6 source that settings tab is labelled Custom Model. Fill in Display Name, API Format, API Base URL and API Key, then click Add Model for the Model Name and Model ID, and finally click Test Connection.
Does the LobsterAI Base URL need /v1?
It depends on the format: no /v1 for Anthropic Compatible, so enter https://api.qcode.cc/api and Test Connection appends /v1/messages; /v1 for OpenAI Compatible, so enter https://api.qcode.cc/openai/v1 and Test Connection appends /chat/completions. The on-screen hint says OpenAI-compatible URLs usually end with /v1, and the official example buttons pair OpenAI Compatible with https://api.siliconflow.cn/v1 and Anthropic Compatible with https://api.siliconflow.cn. Stop the address at /api or /openai/v1, since LobsterAI adds /messages and /chat/completions itself; OpenAI Compatible (Responses) also takes https://api.qcode.cc/openai/v1.
Can LobsterAI use Claude?
LobsterAI's custom models support Anthropic Compatible (Messages), so you can enter QCode's Anthropic address https://api.qcode.cc/api and call Claude models such as claude-sonnet-5-5, claude-opus-5-5 and claude-haiku-5-5. Claude models are served only on that address; under an OpenAI format QCode answers model_not_available_on_endpoint, and switching that provider's API Format back to Anthropic Compatible fixes it. GPT models use the OpenAI address https://api.qcode.cc/openai/v1.
Does LobsterAI support OpenAI Responses?
Yes. The official 2026.10.23 changelog says custom models support OpenAI Responses and Gemini native APIs, and the option is OpenAI Compatible (Responses). For QCode, set Base URL to https://api.qcode.cc/openai/v1; the request lands on /openai/v1/responses, which the endpoints page lists as equivalent to /openai/responses. It serves GPT models such as gpt-6.1-sol and gpt-6.1-terra, and the Test Connection result is the final word. For everyday chat, OpenAI Compatible (Chat Completions) is enough.
Why do the English and Chinese UIs list different built-in providers?
The built-in list follows the UI language: in the 2026.11.6 source the Chinese UI shows the built-in providers tagged china and the English UI shows another list; the maintainer also explains in issues #87 and #185 that the English UI adds built-in entries such as OpenAI and Gemini. Custom providers appear in both languages, so the Custom Model steps for QCode are the same in the Chinese and English UI, and both Claude and GPT come in through custom providers.
Can one QCode key work in both LobsterAI and Claude Code?
Yes. The QCode endpoints page says a single API key works across all three protocols, and the request path chooses the protocol. In Claude Code set ANTHROPIC_BASE_URL to https://api.qcode.cc/api and put the same cr_ key in ANTHROPIC_AUTH_TOKEN; in LobsterAI, Claude uses that same address and GPT uses https://api.qcode.cc/openai/v1. All usage counts on that one key, billed per token, with per-model rates on /models.
Official sources
NetEase Youdao's LobsterAI site and FAQ, the changelog (v2026.8.14, v2026.10.23), the user manual and the model configuration guide; the README and issues #87, #185 and #362 in the GitHub repository netease-youdao/LobsterAI, and the source under the 2026.11.6 tag (UI strings, the Test Connection path, the example buttons and the built-in provider lists); Zhipu's official AutoClaw docs and blog. All read on 2026-11-10. QCode: the endpoints page, the LobsterAI, OpenClaw and AutoClaw docs, and /models (2026-11-10).
Connect LobsterAI with one QCode key
Put QCode into a LobsterAI custom provider: Anthropic Compatible for Claude, OpenAI Compatible for GPT, and either for Chinese models; billed per token, with per-model rates on /models.
Related reading
Pay-as-you-go Packs
One-time quota packs valid for 7 days, no subscription; extra packs add to the same key
ClawHub (Oct 2026): OpenClaw Skills & Plugins Registry
As of 2026-10-08, ClawHub is the public registry for OpenClaw skills and plugins: install with openclaw commands, publish with the clawhub CLI, auto-scanned.
OpenClaw on a cloud server
Checked in October 2026: putting the OpenClaw Gateway on a cloud server, the background service and the lingering step, the trial and the terms, the renewal rules and the QCode endpoint setup.
Crypto payment guide
Nine coins, 56 networks, zero platform fee for AI coding API credits.
Statements on this page were checked on 2026-11-10 against the official LobsterAI README, user manual and changelog and the 2026.11.6 source on GitHub; the official pages prevail. Menu names follow the version you install, QCode models follow /models, and the Test Connection result confirms the address and key.