China AI Hub China AI Hub

Agents / Qoder

Qoder

Commercial agentic coding platform ('Qoder - The Agentic Platform') with desktop app (Qoder IDE / Qoder), CLI (qodercli), JetBrains plugin, Cloud Agents API, and work agents (QoderWork, QoderWake). Agentic loop of understand-plan-execute-verify-iterate with plan- or goal-driven workflows, Expert team multi-agent mode, built-in browser, Memory and Knowledge Base, scheduled automations and enterprise governance. Closed-source; presented under the Alibaba Cloud Model Studio ecosystem.

Qoder is a platform agent by alibaba-cloud, built on qwen3.8-max, qwen3.8-flash, deepseek-v4-pro, deepseek-v4-1-flash, glm-5.3, glm-5.3-flash, kimi-k3, minimax-m3, deployable as both a cloud service and self-hosted.

Qoder
Image: AI-generated illustration (Seedream)
Last verified: · Review cadence: 90 days · Next review:
Key facts for Qoder
Key factValue
Companyalibaba-cloud
Typeplatform
FrameworkCommercial desktop/CLI platform; Auto tier smart-routes tasks across selectable models (Ultimate ~1.6x / Performance ~1.1x / Efficient ~0.3x credit multipliers)
Deploymentboth
Open sourceNo (proprietary)
Documentationhttps://docs.qoder.com/qoder/overview.md
PricingInternational (qoder.com): Free $0 (one-time 2-week Pro trial, 300 Credits); Pro $20/mo (4,000 Credits); Pro+ $60/mo (6,000); Ultra $200/mo (20,000); Credit Pack $20/1,500 (1-month validity). Enterprise via contact sales. China version (qoder.cn) billing via Alibaba Cloud plans - CN pricing not publicly disclosed.

Capabilities

Capabilities of Qoder
CapabilitySupported
Tool callingYes
Browser useYes
Computer useYes
MCP supportYes
MemoryYes
PlanningYes
Multi-agentYes
APIYes

Use cases

Limitations

Short answer. Qoder is Alibaba Cloud’s commercial agentic coding platform — desktop app (Qoder IDE), CLI (qodercli), JetBrains plugin, Cloud Agents API and work agents (QoderWork, QoderWake) — with plan- or goal-driven workflows and a multi-agent Expert team mode. It is Alibaba’s closed, enterprise-governance coding surface, distinct from the open Qwen Code and Qwen-Agent.

Key facts.

  • Its Auto tier smart-routes tasks across multiple vendors’ models — Qwen3.8-Max, DeepSeek-V4-Pro, GLM-5.3, Kimi K3, MiniMax-M3 — with Ultimate/Performance/Efficient credit multipliers of ~1.6x/1.1x/0.3x.
  • An agentic loop of understand-plan-execute-verify-iterate, with built-in browser, Memory and Knowledge Base, scheduled automations and enterprise governance.
  • International pricing (qoder.com): Free $0 (one-time 2-week Pro trial, 300 Credits); Pro $20/mo (4,000 Credits); Pro+ $60/mo (6,000); Ultra $200/mo (20,000); Credit Packs $20/1,500. China (qoder.cn) via Alibaba Cloud plans, CN pricing undisclosed.
  • Closed-source: only SDKs, changelogs and skills are public; no public GitHub repository (verified against XGenerationLab and Qoder-AI on 2026-09-27).
  • Operating entity listed as BRIGHT ZENITH PRIVATE LIMITED; its corporate ownership relationship to Alibaba is not publicly disclosed.

What this means. Qoder’s value is routing and governance, not any single model’s capability — it deliberately spans competitors’ models by credit tier, making it a managed multi-model orchestration layer rather than a model-bound agent.

What is uncertain. The exact launch date is not disclosed (IDE release notes start 2025-08-21), China pricing is not public, and the Alibaba ownership relationship is not publicly documented — the “Alibaba Cloud” association rests on the Model Studio integration guide, not a disclosed corporate link.

Sources.

evidence_idsource_namesource_urlsource_typepublishedverifiedconfidenceconflict
src-agents-qoder-1Qoder official sitehttps://qoder.com/Official—2026-09-20high—
src-agents-qoder-2Qoder docs - overviewhttps://docs.qoder.com/qoder/overview.mdOfficial documentation—2026-09-20high—
src-agents-qoder-3Qoder pricinghttps://qoder.com/pricingOfficial documentation—2026-09-20high—
src-agents-qoder-4Alibaba Cloud Model Studio Qoder integration guidehttps://www.alibabacloud.com/help/en/model-studio/qoder-agentOfficial documentation—2026-09-20high—

Why it matters

Qoder is the multi-model aggregator in Alibaba’s coding portfolio, and the database’s clearest signal that the Chinese agent market has matured to an enterprise tier where the differentiator shifts from raw model access to managed orchestration. Its Auto tier routes work across Qwen, DeepSeek, GLM, Kimi and MiniMax models by credit multiplier — a selection layer, not a binding to any one model.

China AI Hub analysis indicates Qoder matters as evidence of the enterprise tier’s logic: a closed platform that earns its margin from routing, governance, scheduled automation and team management — not from selling a model it built. The 1.6x/1.1x/0.3x credit multipliers are a pricing-encoded quality ladder across rival vendors’ models, which is only coherent if Qoder’s value proposition is “we pick the right model per task,” a genuinely different bet from Qwen Code’s “you pick.”

How it differs from Qwen Code and Qwen-Agent

Alibaba fields three coding/agent surfaces that split cleanly on openness and model posture.

  • Qwen Code is the open Apache-2.0 multi-protocol coding agent — a free CLI/desktop/browser surface where the user brings any model (Qwen, DeepSeek, Kimi, MiniMax, OpenAI, Anthropic, Gemini, local). Model-agnostic and open-source.
  • Qwen-Agent is the open Apache-2.0 Python framework — the low-level, Qwen-native plumbing (Assistant/FnCallAgent/ReActChat, @register_tool) that Qwen Code and Qwen Chat build on.
  • Qoder is the closed commercial platform — the governance layer with Auto-tier routing, Expert teams, Memory/Knowledge Base, scheduled automations and enterprise controls, presented under the Model Studio ecosystem.

China AI Hub analysis: the three are a ladder — Qwen-Agent is the framework, Qwen Code is the open client, Qoder is the managed product — and the openness declines as the governance rises. Qwen-Agent and Qwen Code are inspectable and free; Qoder is closed and billed, trading transparency for routing and controls. A self-hosting developer uses the open pair; an enterprise buying governance and multi-model routing buys Qoder.

Practical implications

For enterprises. Qoder is the governance path — Expert team multi-agent mode, Memory and Knowledge Base, scheduled automations, cloud agents via API, and a 500-tool-round-per-task ceiling (IDE v1.28.0) that bounds runaway tasks. The cost is vendor opacity: the CLI/IDE source is closed, only SDKs/changelogs/skills are public, and the Alibaba ownership relationship is not disclosed.

For cost management. The credit system is strict — unused monthly credits expire, no refunds after 24h or after credits are used, and the free tier has limited completions with a one-time Pro trial unavailable on VMs. Teams must plan credit consumption deliberately.

For evaluators. The Alibaba association should be verified directly: the operating entity is BRIGHT ZENITH PRIVATE LIMITED, and the “Alibaba Cloud” link rests on a Model Studio integration guide rather than a disclosed corporate relationship.

What the evidence shows

The evidence is strong on the product and pricing, and honest about the corporate opacity. The docs document the agentic loop, Auto-tier routing, Expert teams, Memory/Knowledge Base, scheduled automations and the multi-surface distribution (IDE, CLI, JetBrains, Cloud API, work agents); the pricing page fixes the international tiers ($20/$60/$200 plus Credit Packs); the Model Studio guide establishes the Alibaba Cloud ecosystem presentation.

The gaps are corporate and historical. The exact launch date is not disclosed (release notes start 2025-08-21), China pricing is not public, and the Alibaba ownership relationship is not documented — the operating entity is BRIGHT ZENITH PRIVATE LIMITED. China AI Hub analysis indicates this opacity is a known feature of the enterprise-closed tier: Qoder’s value is governance and routing, which it can sell without disclosing its corporate structure — but a buyer whose procurement requires knowing the vendor’s ownership should confirm the Alibaba relationship directly rather than infer it from the ecosystem positioning.

Where this fits

WorkloadRelevance
End-to-end delegated coding (Quest mode)High
Multi-model routing by credit tier (Auto)High
Enterprise governance and Expert teamsHigh
Scheduled automations and CI/CD (headless CLI)High
Cloud agents via APIHigh
Self-hosted / open-source inspectionLow (closed-source)
Independent verification of Alibaba ownershipNot publicly documented

Relevance judgments are China AI Hub analysis based on documented capabilities, not vendor claims.

Field reference

FieldValueEvidence type
Underlying modelQwen3.8-Max, Qwen3.8-Flash, DeepSeek-V4-Pro, DeepSeek-V4.1-Flash, GLM-5.3, GLM-5.3-Flash, Kimi K3, MiniMax-M3Official
Target usersDevelopers and enterprises (commercial platform)Official
PlatformDesktop app (Qoder IDE), CLI (qodercli), JetBrains plugin, Cloud Agents APIOfficial
OSNot publicly documentedNot publicly documented
Browser / computer useBuilt-in browser and Computer UseVendor-reported
CodingYes (agentic coding platform)Official
Autonomous task executionYes (plan/goal-driven, Expert team multi-agent)Vendor-reported
MCPYesOfficial
Tool callingYesOfficial
MemoryYes (Memory and Knowledge Base)Official
WorkflowUnderstand-plan-execute-verify-iterate; Auto tier smart-routes modelsOfficial
APIYes (Cloud Agents API)Official
PricingInternational $20/$60/$200 per month; China via Alibaba CloudVendor-reported
RegionInternational (qoder.com) and China (qoder.cn)Official
Open-sourceNo — proprietaryOfficial
DeploymentCloud and self-hostedOfficial
LimitationsClosed-source; 500 tool rounds/task; operating entity not Alibaba-disclosedOfficial
SourceQoder docsOfficial
Last verified2026-09-20Official

See the Alibaba Cloud company profile, the open Qwen Code and Qwen-Agent counterparts, the choosing-a-coding-model guide, and the site’s AI agents and tool calling technology pages.

Labels used above: Official fact (from the Qoder site, docs and the Alibaba Cloud Model Studio integration guide), Vendor-reported claim (pricing and credit-multiplier statements by Qoder), and China AI Hub analysis (our synthesis, always introduced as such).

Underlying models

Sources

Data interpretation

Evidence types for Qoder
FieldValueEvidence type
Companyalibaba-cloudOfficial
TypeplatformOfficial
Underlying modelsqwen3.8-max, qwen3.8-flash, deepseek-v4-pro, deepseek-v4-1-flash, glm-5.3, glm-5.3-flash, kimi-k3, minimax-m3Official
Open sourceNo (proprietary)Official
DeploymentbothOfficial
CapabilitiesTool calling, Browser use, Computer use, MCP support, Memory, Planning, Multi-agent, APIOfficial

Evidence types: Official = vendor documentation or official pages. See the sourcing policy.

What is Qoder?

Commercial agentic coding platform ('Qoder - The Agentic Platform') with desktop app (Qoder IDE / Qoder), CLI (qodercli), JetBrains plugin, Cloud Agents API, and work agents (QoderWork, QoderWake). Agentic loop of understand-plan-execute-verify-iterate with plan- or goal-driven workflows, Expert team multi-agent mode, built-in browser, Memory and Knowledge Base, scheduled automations and enterprise governance. Closed-source; presented under the Alibaba Cloud Model Studio ecosystem.

Which models does Qoder use?

Qoder uses qwen3.8-max, qwen3.8-flash, deepseek-v4-pro, deepseek-v4-1-flash, glm-5.3, glm-5.3-flash, kimi-k3, minimax-m3.

Is Qoder open source?

No — Qoder is not open source.

How much does Qoder cost?

International (qoder.com): Free $0 (one-time 2-week Pro trial, 300 Credits); Pro $20/mo (4,000 Credits); Pro+ $60/mo (6,000); Ultra $200/mo (20,000); Credit Pack $20/1,500 (1-month validity). Enterprise via contact sales. China version (qoder.cn) billing via Alibaba Cloud plans - CN pricing not publicly disclosed.

Where does China AI Hub get its Qoder data?

From 4 sources, last verified 2026-09-20.