Latest update (Oct 15, 2026): The open-source coding-agent wave keeps widening. APUS AI open-sourced Reflex Agent (Oct 13, MIT license) — a Personal Agent built on its open Reflex decision model plus a Continuum memory engine — claiming response speeds up to 3x faster than Claude Code and Codex in tests, with “infinite context” powered by lossless external memory instead of compressed summaries. The same week, OpenCode 2.0 shipped (Oct 13): a single background service shared by terminal, web and desktop clients, adding Code Mode, background subagents, 80+ plugins and an embeddable SDK (MIT). For developers picking an AI coding platform, agent engineering and open ecosystems are becoming decisive — Alibaba’s Qoder competes head-on with 40+ connectors, 70+ plugins and 20,000+ skills, and its free Community Edition (BYOK) remains a zero-cost way to try agentic coding.
📰 Latest update (Oct 10, 2026): Lenovo’s in-house coding-agent framework TianxiCode, paired with DeepSeek-v4.1-Flash, took #1 on SWE-bench-Live (Lite tier) with a 71% problem-resolution rate and passed the benchmark’s official “Verified” audit (Oct 9) — the first verified top ranking by a Chinese-built coding agent on the toughest real-world software-engineering test. It underscores that agent engineering (framework + workflow) is the new battleground on top of strong models — exactly where Alibaba’s Qoder and Qwen Code compete, and where Qoder’s free Community Edition (BYOK) lets you try agentic coding today.
📰 Latest update (September 2026): At Apsara Conference 2026 (Sep 22-24, Hangzhou), Lenovo Tianxi AI 4.3 debuted with its on-device agent framework Tianxi AgentCore and announced a partnership with Alibaba Qoder to improve user experience together, accelerating the blend of on-device agents with cloud coding workbenches. Meanwhile, the Qwen AI platform added Agent Studio and API Express Mode (1.5-2x TPS); as Alibaba Cloud's agentic coding platform, Qoder is now more tightly connected to enterprise agent services and industry solutions. (Reported by ZOL and QbitAI, Sep 23, 2026.)
📰 Latest update (August 2026): Qoder launched its new desktop app with a voice-enabled desktop agent you can talk to anytime — it plans tasks, writes code, and verifies results itself via Computer Use / Browser Use. It now supports 40+ connectors, 70+ plugins, and 20,000+ skills, serving 6 million users and 100,000+ enterprises. During the launch event, individual users can claim 500 Qwen 3.8-dedicated Credits daily. (Reported by QbitAI, August 29, 2026.)Latest (Sep 26): Microsoft rolled out its revamped Copilot app, positioning it as a “super app” — the Cowork, Code and Autopilot modules switch to usage-based pricing (the traditional Microsoft 365 Copilot commercial subscription is $30/user/month). The new Code capability builds on GitHub Copilot’s underlying tech so non-programmers can create simple software in natural language; fewer than 7% of the 450M+ Office 365 commercial seats have purchased AI add-ons so far. The enterprise AI assistant market is shifting from subscriptions toward pay-as-you-go — converging with Qoder’s Credits model. For Chinese developers and enterprises, Qoder CN (fully onshore models + data sovereignty) remains the differentiated choice. (Reported by Sina Finance et al., Sep 25, 2026.)
1. What Is Qoder?
Qoder is Alibaba Cloud's agentic coding platform, first released in August 2025. On August 27, 2026 — its first anniversary — Qoder received a major upgrade, repositioning from an "AI coding tool" to an "agentic workbench for everyone." At launch, Qoder had accumulated over 6 million global users and 100,000+ enterprise customers (per Alibaba Cloud's August 2026 anniversary announcement).
There are two editions:
- Qoder International (qoder.com): For global users, supports Claude, GPT, Gemini, Qwen, and other models via Alibaba Cloud Model Studio.
- Qoder CN (qoder.com.cn, formerly Tongyi Lingma): For the China market, powered by Chinese LLMs (Qwen, GLM, DeepSeek, Kimi, MiniMax), with the full stack deployed on Chinese cloud infrastructure for data sovereignty in finance and government.
2. What's New in the August 27 Upgrade?
This isn't a feature iteration — it's a fundamental repositioning:
Dual Modes: Coding + General
Coding Mode gives developers the full code workflow — completion, Agent chat, Quest autonomous tasks, and RepoWiki documentation. General Mode hides the code layer entirely. Non-technical users describe what they want in natural language, and Qoder handles data processing, file operations, web browsing, and desktop app control automatically.
Quest Autonomous Tasks
Quest is Qoder's core agent capability. Give it a goal and it: understands context → makes a plan → calls tools → executes → verifies results → self-corrects. It can search codebases, edit across files, run tests, and iterate based on real output. Complex multi-step development tasks run asynchronously, and you can inspect, adjust, or take over at any time.
Desktop Pet + Voice Interaction
The new desktop pet sits on your desktop. Click it and speak your request; it tracks task progress in real time and proactively pushes reminders and next-step suggestions. No need to switch to the IDE window.
Auto Model Scheduling
Non-experts don't need to pick models. The Auto engine balances output quality, speed, and cost based on task type and complexity. Power users can still freely switch between Claude, GPT, Gemini, and Qwen.
40+ Connectors, 70+ Plugins, 20K+ Skills
Qoder integrates with code repositories (GitHub/GitLab), project management (Jira, etc.), cloud services (Alibaba Cloud, etc.), and internal enterprise systems. MCP support allows custom tools and skills.
Long-term Memory & Skill Distillation
Cross-session memory remembers your project structure, coding style, and preferences. It periodically distills reusable skills from interactions — the more you use it, the better it understands your workflow.
Qoder CN Full Suite
The China edition now spans a complete product matrix: one account for IDE, JetBrains plugin, CLI terminal, QoderWork desktop assistant, QoderWake digital employees, mobile app, and Cloud Agents — with shared Credits across all products.
3. Complete Pricing
Qoder International (Token Plan, USD)
| Plan | Price | 7-Day Credits | Concurrent Agents |
|---|---|---|---|
| Community (Free) | $0 | Core features, BYOK supported | — |
| Pro | $20/mo | 2,000 Credits/mo | Up to 5 |
| Pro+ | $60/mo | 6,000 Credits/mo | Higher |
| Ultra | $200/mo | 20,000 Credits/mo | Maximum |
| Note: a limited-time 50%-off launch promo (Pro $10 / Pro+ $30 / Ultra $100) ended April 30, 2026; standard pricing shown above. | |||
Team Edition (USD)
| Seat | Price | Monthly Credits |
|---|---|---|
| Teams | $40/seat/mo | 3,000 Credits/seat per month |
| Enterprise | Custom | SSO, audit logs, dedicated support — contact Qoder sales |
Qoder CN Domestic Pricing (CNY)
| Plan | Price | Credits | Key Features |
|---|---|---|---|
| Community | Free | Limited completion + chat | Basic code completion |
| Personal Pro | ¥59/mo | 2,000/mo | Unlimited completion, RepoWiki, Quest, Experts |
| Personal Pro+ | ¥169/mo | 6,000/mo | All Pro features + higher quota |
| Enterprise Teams | ¥99/seat/mo | 3,000/seat/mo | Enterprise knowledge base, enhanced codebase |
| Enterprise VPC | ¥199/seat/mo | 3,000/seat/mo | VPC deployment, IP whitelist, private network |
Save More with Off-Peak Discounts
Qoder CN offers time-based Credits discounts on premium models (Beijing time 22:00–08:00), up to 80% off:
| Model | Standard Rate | Off-Peak (22:00–08:00) |
|---|---|---|
| Qwen3.8-Max | 0.5x | 0.25x (50% off) |
| Qwen3.7-Max | 0.5x | 0.1x (80% off) |
| Qwen3.7-Plus | 0.1x | 0.04x (60% off) |
Discounts apply automatically on weekends and holidays for Pro Trial, Pro, and Teams users.
4. Qoder vs Cursor vs GitHub Copilot vs Windsurf
| Feature | Qoder | Cursor | GitHub Copilot | Windsurf |
|---|---|---|---|---|
| Vendor | Alibaba Cloud | Anysphere | Microsoft/GitHub | Cognition Labs |
| Autonomous mode | Quest | Agent mode | Workspace Agent | Cascade |
| Multi-model | Claude/GPT/Gemini/Qwen | Multi-model | GPT/Claude | Multi-model |
| Non-coder General mode | Yes | — | — | — |
| Desktop pet / voice | Yes | — | — | — |
| Free tier | $0 + BYOK | Limited free | Free for students/OSS | Limited free |
| Paid starts at | $20/mo or ¥59/mo | $20/mo | $10/mo | $15/mo |
| JetBrains plugin | Yes | VS Code focus | All major IDEs | VS Code focus |
| China compliance | Qoder CN onshore | — | — | — |
| 24/7 digital employees | QoderWake | — | — | — |
Qoder differentiates with its General mode for non-programmers, Qoder CN for Chinese data sovereignty, QoderWake for 24/7 autonomous operation, and deep Alibaba Cloud ecosystem integration.
5. Use Cases
Developers
- Full-feature development: Quest mode breaks down requirements, writes code, runs tests, and fixes bugs autonomously
- Code understanding: RepoWiki auto-generates project docs (up to 6,000 files, Git repos only)
- Model flexibility: Claude for architecture, Qwen for Chinese tasks, GPT for code generation
- MCP extensibility: Connect databases, CI/CD, monitoring, and external tools
Non-technical Users (General Mode)
- Batch file processing: Image compression, format conversion, renaming, watermarking — describe in plain language, get a desktop app
- Data analysis: Upload Excel/CSV, ask questions in natural language, get charts and reports
- Web automation: Browser Use can operate websites, fill forms, and scrape data
- Office automation: Connect to Office, email, calendar via built-in connectors
Enterprise Teams
- QoderWake digital employees: 24/7 autonomous agents for code review, bug diagnosis, data processing
- Enterprise knowledge base: Private codebase and document enhancement with shared context
- VPC deployment: Dedicated network for finance, government, and regulated industries
- Multi-seat management: Usage analytics, access control, audit logs
6. How to Get Started
- Download: International at qoder.com/download, China at qoder.com.cn/download. Supports Windows 10/11, macOS 11+, Linux.
- Sign up: New users automatically get a 14-day Pro trial with 300 Credits; after the trial, the Pro plan is $20/month (2,000 Credits).
- Configure model access: In Settings → Models → Add, select "Alibaba Cloud Model Studio - International" (or China for Qoder CN), choose Token Plan, Coding Plan, or pay-as-you-go.
- Start using: Coding mode works directly in the IDE; switch to General mode to describe tasks in natural language.
- CLI install (optional): Run
curl -fsSL https://qoder.com/install | bashin your terminal.
Start with Qoder + Alibaba Cloud Free Credits
New users get a 14-day Pro trial with 300 Credits. Alibaba Cloud also offers $200 in free credits for new international accounts.
Get $200 Free Credits → ECS from $4.50/mo →📰 Market signal, Sept 8, 2026: Cognition hits a $48B valuation
On September 8, 2026, Cognition — the company behind the Devin autonomous coding agent and the Windsurf IDE — closed an oversubscribed Series E above $2B at a $48 billion valuation, nearly doubling the $26B mark from its May 2026 Series D in just four months. Run-rate revenue climbed from $492M to nearly $900M over the same window. Andreessen Horowitz and Accel led the round; NVIDIA is among the participants. Devin is now used by engineering teams at NVIDIA, GE Aerospace, Citi, Mercedes-Benz, and Modal. The new product surfaces — Devin Auto-Triage, Devin Security Swarm, and Devin Automations — push the agent from chat-style assistance toward an autonomous operations layer triggered by events in Slack, GitHub, and Linear.
For Alibaba's Qoder — and any agentic workbench competing in this space — the takeaway is unambiguous: enterprises are willing to pay real money for autonomous coding that goes beyond autocomplete, and the AI coding market is far from winner-take-all. To stay competitive over the next 12 months, Qoder will need to ship deeper agent autonomy, broader enterprise integrations, and a credible monetization story on par with Cognition's $900M run-rate trajectory.
Sources: Unite.AI — Cognition Raises Over $2B at $48B Valuation; PYMNTS — Cognition Secures $48B Valuation. Original pricing, feature comparison and positioning analysis above are unchanged; this card only adds the latest market benchmark.
📰 Market signal, Sept 17, 2026: Claude Code relaunches Projects for multi-agent coding
On September 17, 2026, Anthropic rolled out a redesigned Projects experience in Claude Code (beta): instead of manually splitting work across sessions, you describe the goal and Claude scopes, delegates and coordinates parallel threads - each a full cloud Claude Code session working on its own branch and repo copy - backed by shared memory and an artifact library. Overlaps between threads are resolved as standard merge conflicts (PRs), and work continues after you step away from the computer. Beta access starts with select Claude Pro and Max subscribers and will widen over coming weeks to Team and Enterprise plans.
The takeaway for Qoder - and every agentic workbench - is clear: the AI coding race is shifting from autocomplete to multi-agent orchestration and autonomous project management. Whoever nails the close-the-loop experience of goal, parallel execution, conflict resolution and delivered result at controlled cost becomes the default entry point for engineering teams. Qoder's Quest autonomous tasks and Auto scheduling engine already target the same direction, and its cloud-elastic-compute-plus-workbench integration remains a differentiator.
Sources: Anthropic - Projects redesigned: from folder to conversation; The Verge. Original pricing, feature comparison and positioning analysis above are unchanged; this card only adds the latest market benchmark.
📰 Market signal, Sept 19, 2026: Claude Code adopts AGENTS.md - cross-tool project rules become standard
On September 18-19, 2026, Anthropic added AGENTS.md support to Claude Code 2.1.277: when a project folder has no CLAUDE.md, Claude Code now reads AGENTS.md instead (toggleable in /config; not yet on Bedrock, Vertex or Foundry). CLAUDE.md is Claude Code-specific workflow/MCP/command configuration, while AGENTS.md is the shared, cross-tool project rules file already adopted across Cursor, Codex, Copilot and other ecosystems. With this move, "one instruction file for every AI tool" is becoming the industry default - developers switching tools, onboarding to a repo or contributing to open source no longer have to re-configure project rules for each tool.
The takeaway for Qoder - and every agentic workbench: AI coding tools are moving from private configuration to open rule interoperability. AGENTS.md is fast becoming the de facto project-rules standard; workbenches that embrace open standards and lower switching/collaboration costs are more likely to become the default entry point for engineering teams. That is exactly the direction Qoder's open ecosystem (20K+ skills, 40+ connectors) keeps pushing.
Sources: Claude Code Changelog (2.1.277); Hacker News discussion. Original pricing, feature comparison and positioning analysis above are unchanged; this card only adds the latest market benchmark.
7. FAQ
8. Verdict
Qoder's one-year anniversary upgrade marks a substantive shift from AI coding tools toward "universal agentic workbenches." Its core value isn't just stronger coding — it's opening Agentic Coding capabilities to people who don't write code. For developers, Quest mode and multi-model support match or exceed Cursor's autonomous development. For enterprises, Qoder CN's compliance deployment and QoderWake digital employees offer a unique productivity solution.
The main caveat is Credits consumption — Quest mode costs ~50 Credits per request, and long tasks can be a "Credits black hole." We recommend thoroughly testing your usage patterns during the free trial before committing. Overall, the $0 free tier with BYOK makes entry cost-free, and the $20/month Pro pricing (2,000 Credits) is competitive among alternatives.