Tencent CodeBuddy
Tencent Cloud's coding assistant (腾讯云代码助手), a distinct product from the Hunyuan chatbot. Standalone IDE, editor plugins and a CLI with native Agent Client Protocol support.
Overview
This is a separate product from Tencent's Hunyuan model line, and the distinction gets lost constantly in English coverage. Hunyuan is the model family and the chatbot. CodeBuddy is Tencent Cloud's coding assistant, which consumes Hunyuan output among other models. The internal number is the one worth knowing: Tencent reports that more than half its own R&D staff use it, which across a codebase of that size is a harder signal than any download count.
CodeBuddy Code 2.0 shipped in January 2026 and is the piece that deserves attention. It added Plan Mode, an isolated sandbox for executing generated code, custom sub-agents, Hooks, SDK integration and a plugin marketplace. It also speaks the Agent Client Protocol natively, the standard Zed introduced to decouple agent servers from editor clients, so CodeBuddy can serve any ACP-compatible editor rather than binding you to Tencent's own. Tencent covered the full range of form factors ahead of its domestic competitors: a standalone AI code editor, VS Code and JetBrains plugins, and a CLI.
The model story splits by region and runs opposite to what most people assume. The international build integrates Claude, GPT and Gemini. The mainland build runs Hunyuan and DeepSeek. Tencent's own Hy4 preview, 770 billion total parameters with 49 billion active and a 1M token window, went into both after being open-sourced. International pricing is genuinely low: free at 100 credits a month, Pro at $10 a month or $96 a year for 2,000 credits, Team at $40 per seat. The sting is top-ups, where $15 buys 500 credits that expire in a month. Registration uses Google or GitHub, so no mainland phone number is required. English documentation covers the product but reads as translation, and announcements reach Chinese channels first.
Key Features
- ✓ CodeBuddy Code 2.0, shipped January 2026, adding Plan Mode, an isolated sandbox for code execution, custom sub-agents, Hooks, SDK integration and a plugin marketplace
- ✓ Native Agent Client Protocol support, so it registers as an agent server inside Zed or any other ACP-compatible client rather than requiring Tencent's own editor
- ✓ Full set of form factors: standalone AI code editor, VS Code and JetBrains plugins, and a terminal CLI, the first from a Chinese vendor to cover all three
- ✓ International build integrates Claude, GPT and Gemini; the mainland build runs Tencent Hunyuan and DeepSeek
- ✓ Hunyuan Hy4 preview, at 770 billion total parameters with 49 billion active and a 1M token context window, rolled into both the domestic and international products
- ✓ Sign-in by Google or GitHub authorisation on the international site, with no Chinese mobile number in the registration path
Where it holds
- • Pro at $10 a month for 2,000 credits undercuts essentially every Western equivalent at the same capability tier
- • ACP support means the agent is portable. You can run it inside Zed and swap it out later without rewriting your workflow
- • Genuine internal validation rather than marketing: Tencent reports more than half its own R&D staff use it, across a codebase few outside companies can match for scale
- • Registration works with a Google or GitHub account, so the international version is usable from anywhere without a mainland cloud identity
Where it breaks
- • Credit accounting is opaque and top-ups punish you, at $15 for 500 credits that expire after a single month
- • The mainland and international builds have different model catalogues, separate accounts and separate billing, so evaluations done on one do not transfer to the other
- • Proprietary throughout. Hy4 weights being open-sourced does not make CodeBuddy itself inspectable or self-hostable
- • English documentation exists and is reasonably complete but reads as translated from the Chinese originals, and new features land on mainland channels before the international docs catch up
My Take
The regional split catches people out: the international build routes to Claude, GPT and Gemini while the mainland build runs Hunyuan and DeepSeek. CodeBuddy Code 2.0 landed in January 2026 with Plan Mode, a sandboxed execution environment and native Agent Client Protocol support, so it slots into Zed as an agent server instead of trapping you in Tencent's editor. Pro at $10 a month for 2,000 credits undercuts almost everything Western, though the $15 top-up for 500 credits that expire in a month is where the margin hides.
Quick Info
- Pricing:
- freemium
- Openness:
- Proprietary
- Starting at:
- International pricing in USD: Free at 100 credits per month, Pro at $10 per month or $96 per year for 2,000 credits (1,000 base plus 1,000 bonus), Team at $40 per seat per month or $480 per seat per year for 1,000 credits per seat drawn from a shared pool. Top-ups are the expensive part, at $15 for 500 credits valid one month for Pro users, with team packs from $40 to $4,000 carrying six-month validity. The mainland Chinese version is billed separately and has historically offered unlimited use of the domestic models.
- Added:
- Aug 2026
- Updated:
- Aug 2026
Use Cases
Judge it on your own work
The notes above say where Tencent CodeBuddy holds and where it breaks. The fastest check is your own workload.
Visit website ↗Alternatives to Tencent CodeBuddy
Qoder
freemiumAlibaba's agentic IDE, which absorbed Tongyi Lingma (通义灵码) as Qoder CN on 20 May 2026. Repo Wiki and Quest Mode on credit-based pricing, with Linux builds since version 1.0.
Trae
freemiumByteDance's AI IDE and standalone SOLO agent that scaffolds projects, runs terminal commands and tests in a browser, at $10 a month for Pro against Cursor's $20.
Cursor
freemiumA VS Code fork built around agentic editing, with its own Composer models, a pull request review bot and SOC 2 controls for teams that want AI in the editor.
GitHub Copilot
freemiumMicrosoft's AI coding assistant across VS Code, JetBrains and the GitHub web, with background agents that open pull requests and org-level policy controls.