# AI apps around the world: connection guides by region

Explore AI apps in India, China, Europe and other markets. Find documented Pear MCP setup paths and understand which consumer apps still need a supported client.

Canonical page: https://pearmcp.com/ai-apps

Documentation checked: 2026-09-26. Published by Pear. This is a guide to documented connection paths, not a national market-share ranking or proof of availability in every country. Network reachability, account and plan access, authentication, and data handling need checking for your location.

## Global clients

- [ChatGPT setup](https://pearmcp.com/guides/chatgpt)
- [Claude setup](https://pearmcp.com/guides/claude)
- [Grok setup](https://pearmcp.com/guides/grok)
- [Perplexity setup](https://pearmcp.com/guides/perplexity)
- [Poke setup](https://pearmcp.com/guides/poke)
- [Raycast setup](https://pearmcp.com/guides/raycast)
- [Gemini CLI setup](https://pearmcp.com/guides/gemini-cli)
- [Muse Code setup](https://pearmcp.com/guides/muse)

A setup guide is not a completed authenticated connection test. Gemini CLI, coding agents and consumer chat apps are separate products. Check the exact guide and its testing status.

## India

Start with the app you already use. Global assistants and India-built products serve different audiences; a local model or language focus does not by itself provide a connection to your accounts.

[Sensor Tower’s February 2026 India report](https://sensortower.com/blog/state-of-india-mobile-app-market-2026) identifies generative AI as a leading source of download growth. It does not establish a ranking for the local products below. Global apps and their supported plans remain relevant wherever available.

### Sarvam Indus

Sarvam introduced Indus in February 2026 as a limited beta built for the Indian context. Its consumer interface is separate from Sarvam’s developer products.

No public Pear setup found for the Indus app. Sarvam publishing its own MCP server does not make Indus a Pear client.

[Official source](https://www.sarvam.ai/blogs/introducing-indus)

### Zoho Zia Chat

Zia Chat advertises connections to external tools through MCP. Its current page describes early access in India and US data centers.

MCP advertised; the reviewed page does not document the remote transport or authentication settings needed for Pear. Setup remains unverified.

[Official source](https://www.zoho.com/ziachat/)

### JioAI Assistant

Jio’s workplace assistant documents its own Microsoft 365 account connection.

A native Microsoft connection is not a custom Pear connection. No user-configurable Pear MCP setup is established by the reviewed account documentation.

[Official source](https://aiassistant.jio.com/docs/account)

## China and Chinese-language AI tools

Choose the exact product: a company’s consumer chat app, coding client and model API can have different connection capabilities. The guides below cover the named coding clients.

### Qwen Code

Qwen Code documents remote HTTP MCP servers with a private Authorization header. Its configuration uses httpUrl.

Setup guide available; authenticated Pear use is still unverified. This is not a setup guide for the consumer Qwen chat app.

[Official source](https://qwenlm.github.io/qwen-code-docs/en/users/features/mcp/) · [Setup and status](https://pearmcp.com/guides/qwen-code)

### Kimi Code

Kimi Code supports remote MCP connections from its coding client and provides a personal MCP configuration.

Setup guide available for Kimi Code. It does not establish MCP support in the consumer Kimi chat app; Pear host testing remains pending.

[Official source](https://www.kimi.com/code/docs/en/kimi-code-cli/customization/mcp.html) · [Setup and status](https://pearmcp.com/guides/kimi-code)

### Qoder Work

Qoder Work has a connector interface for Streamable HTTP servers and request headers.

Setup guide available for the Work desktop app. Qoder CLI has a different configuration schema; authenticated Pear testing remains pending.

[Official source](https://docs.qoder.com/qoderwork/connectors) · [Setup and status](https://pearmcp.com/guides/qoder-work)

### TRAE

TRAE documents manually configured remote MCP servers with a URL and headers.

Setup guide available for the documented IDE interface. Confirm the controls in your installed edition; authenticated Pear testing remains pending.

[Official source](https://docs.trae.cn/ide_add-mcp-servers) · [Setup and status](https://pearmcp.com/guides/trae)

### DeepSeek

DeepSeek offers a consumer assistant and a separate developer API.

No public Pear setup found for the consumer app. Using a DeepSeek model through another MCP-capable client is a separate route.

[Official source](https://api-docs.deepseek.com/news/news250115/)

### Doubao

Doubao is ByteDance’s consumer assistant. It is a different product from the TRAE coding IDE.

No public custom Pear MCP setup found for the consumer app. A TRAE setup guide does not apply to Doubao.

[Official source](https://www.doubao.com/)

### Z.ai / GLM

Z.ai documents an API capability for calling third-party MCP servers over remote transports, including custom request headers.

Developer API route documented; this is not a verified Pear connection in the consumer chat app. A developer must implement and test the agent, account access and approvals.

[Official source](https://docs.z.ai/guides/capabilities/mcp-call)

### Tencent CodeBuddy

CodeBuddy documents MCP connections in its agent SDK, including remote HTTP servers.

Developer SDK route documented. Do not treat SDK configuration as proof that the same controls exist in every CodeBuddy IDE or Tencent Yuanbao interface.

[Official source](https://www.codebuddy.ai/docs/cli/sdk/mcp)

### Cherry Studio

Cherry Studio is a multilingual desktop client with remote MCP support and a choice of model providers.

Remote MCP is documented. Pear-specific authentication and a complete client test are still unverified; check the current server settings against the generic requirements.

[Official source](https://github.com/CherryHQ/cherry-studio-docs/blob/main/advanced-basic/extensions/mcp/README.md) · [MCP requirements](https://pearmcp.com/guides/custom-mcp)

## France and Europe

Mistral’s coding CLI and workplace assistant expose different authentication controls. Choose a guide for the product you are using, rather than applying a legacy Le Chat tutorial to every Mistral interface.

### Mistral Vibe Code

The Vibe Code CLI documents Streamable HTTP and API-key configuration. Its current MCP documentation says OAuth is not supported by the CLI.

Setup guide available using an environment-backed Bearer key. Authenticated Pear host testing remains pending.

[Official source](https://docs.mistral.ai/vibe/code/cli/mcp-servers) · [Setup and status](https://pearmcp.com/guides/mistral-vibe)

### Mistral Work

Work documents custom MCP connectors and detects authentication automatically. It lists Bearer tokens and OAuth with dynamic client registration.

Readiness guide available. Whether Work offers the Bearer path for Pear remains unverified; its documented DCR OAuth flow is not a proven match for Pear’s CIMD flow.

[Official source](https://docs.mistral.ai/vibe/work/connectors/mcp-connectors) · [Setup and status](https://pearmcp.com/guides/mistral-work)

## Japan and South Korea

Localized assistants can connect to selected services without allowing users to add an arbitrary MCP server. Japanese or Korean availability also does not establish where connected data is processed.

### Genspark

Genspark publishes Japanese help content and documents custom or community MCP connections alongside its native integrations. This is a global product, not a claim of Japanese market leadership.

MCP documented; the reviewed help page does not establish the exact Pear transport and Bearer-header setup. Pear compatibility remains unverified.

[Official source](https://www.genspark.ai/ja/helpcenter/connectors-and-integrations)

### Kakao Kanana

Kakao introduced Kanana as a Korean assistant product. Kakao’s wider distribution is not a measurement of Kanana’s own usage.

No public custom Pear MCP setup found in the reviewed product information.

[Official source](https://www.kakaocorp.com/page/detail/11557)

### NAVER AI National Secretary

NAVER announced an assistant for public-service workflows in its app in March 2026.

The announcement describes native service integrations, not user-configured Pear MCP connections.

[Official source](https://navercorp.com/media/pressReleasesDetail?seq=34378)

## Arabic and Russian-language assistants

A regional assistant, its underlying model and its developer platform are separate integration targets. Confirm the client interface before providing a personal API key.

### Jais Chat

Jais Chat is an Arabic-focused assistant associated with MBZUAI, with mobile apps.

No public custom Pear MCP setup found. Access to a Jais model is not evidence of MCP controls in the Jais Chat app.

[Official source](https://www.jaischat.ai/jaischat/en)

### Yandex Alice AI

Yandex offers Alice AI and a separate business product with its own integrations.

The reviewed business announcement does not document adding an arbitrary Pear MCP server to the consumer assistant.

[Official source](https://www.yandex.com/company/news/2026-09-09-01)

### Sber GigaChat

Sber documents building a custom agent that uses a GigaChat model and an MCP client.

Developer-built agent route, not a demonstrated custom MCP setting in the consumer GigaChat app. Pear access and approval behavior still need implementation and testing.

[Official source](https://developers.sber.ru/docs/ru/gigachain/tutorials/agent-gigachat-mcp)

## A model is not an MCP client

If another MCP-capable client supports your model provider and tool calling, configure the model and Pear separately. This does not add MCP to the model vendor’s consumer app. Consider [Jan](https://pearmcp.com/guides/jan), [LM Studio](https://pearmcp.com/guides/lm-studio), [AnythingLLM](https://pearmcp.com/guides/anythingllm), or the [generic requirements](https://pearmcp.com/guides/custom-mcp).

Pear uses remote Streamable HTTP at `https://pearmcp.com/api/mcp`. Follow the client’s documented private API-key or OAuth path. Start with `pear_provider_status`, then a bounded read on a known calendar or message. Review drafts before approving writes. A local app may send provider data to a cloud model; review [data handling](https://pearmcp.com/data-handling).

## Questions

### Can Pear connect to every AI app?

The app must act as a remote Streamable HTTP MCP client and support an authentication path Pear accepts. A model API, native calendar integration or separate MCP server does not establish that capability. A setup guide describes a documented route; a complete authenticated client test is separate evidence.

### Can I use a Chinese or Indian model with Pear?

Potentially, through an MCP-capable client that supports that model provider and tool calling. Configure the model in that client and Pear as a separate MCP server. This does not add MCP support to the model vendor’s consumer app. Check language quality, tool use, data handling and availability for the combination you choose.

### Does a regional listing mean Pear works from that country?

No. These sections describe product and language context, not tested network reachability, local availability or data residency. Client access, plans, model access, provider accounts and local network conditions can differ. Check the current vendor requirements before connecting.

### Which accounts can I connect?

Pear supports iCloud Calendar, Contacts and Mail, and Microsoft 365 Calendar, To Do, Contacts and Outlook Mail. Apple Reminders support is limited to legacy CalDAV data. Google Workspace remains coming soon. Available actions depend on your account, plan and permissions.

## More information

- [All setup guides](https://pearmcp.com/docs)
- [Compatibility evidence](https://pearmcp.com/mcp-compatibility)
- [Poke and assistant alternatives](https://pearmcp.com/ai-assistants)
- [Pear plans](https://pearmcp.com/pricing)

Public X posts are discovery leads, not compatibility evidence. Vendor sources determine setup guidance; missing public instructions are recorded as unknown rather than proof a feature can never work.
