What it does
5 features
Qwen Code · Coding agent
Product and access
Available coding agent from the Qwen team, developed independently.
Core features
5 features from provider sources · checked on 19 October 2026. Availability and limitations follow below.
Usable AI models
Open choice
You use Qwen Code with a Qwen subscription, the Alibaba Coding Plan or your own API keys for OpenAI, Anthropic and Gemini interfaces. An OpenAI-compatible endpoint also lets you use OpenRouter and local servers such as Ollama, LM Studio or vLLM. You can switch models during a session.
Model families named
- Alibaba (Qwen)
- OpenAI (GPT)
- Anthropic (Claude)
- Google (Gemini)
- Z.ai (GLM)
- Moonshot (Kimi)
- MiniMax
- Other models can be connected freely
- Model choice
- You choose
- Your own API key
- Yes
- Router or custom endpoint
- Yes
- Local models
- Yes
The rating covers the generally available main access. Documented · Sources accessed on
Qwen Code: Repository · Qwen Code: Model providers · Qwen Code: Authentication
Does it fit my task?
2 practical tasks · 3 requirements and limits
A good fit if …
- you want a terminal coding agent and choose the model yourself
- you want to build it into scripts or CI through headless calls
- you can configure approval modes and sandbox yourself in settings files
Less suitable if …
- you expect the former free Qwen OAuth tier; it was discontinued on 15 April 2026
- you need central management with roles and audit logs; the reviewed pages describe only system-scope settings
Editorial assessment based on provider documentation. Not a product test.
Practical tasks
Editorial examples based on documented features. No product tests performed.
Example task 1
- Your input
- Fix a bug with a reproducible test.
- Possible result
- A draft for expert review.
- What to check
- Check inputs, output and performed actions.
Example task 2
- Your input
- Explain a module and suggest refactoring.
- Possible result
- A draft for expert review.
- What to check
- Check inputs, output and performed actions.
Fix a bug with a reproducible test.
Your possible prompt
Read the project rules. Investigate [error], propose a small change, and run the relevant checks. Show me the diff and any remaining uncertainties.
Adjust the task, data and permissions.
Requirements and limits
Not yet confirmed
Output quality
Success rates and German-language operation have not been tested here.
Source 1
Can I use it?
With requirements
Europe & regional availability
Installation instructions and model prerequisites reviewed. The assessment covers the local operating path described; complete country and feature availability of all connected services remains open.
Reviewed scope: Local software and documented self-hosting; external services assessed separately
Reviewed on:
With requirements
Self-hosting with own model access
Qwen Code documents local installation for Linux, macOS and Windows and own API providers or local models through Ollama and vLLM. Regional conditions apply to the model service actually selected.
Region, reason and source
Region covered: Europe: own computer or server; model services are country-specific
Documented reason: Technical assessment based on installation and provider selection. These sources do not establish a legal cause for regional restrictions.
Accounts & permissions
Based on provider documentation. Permissions depend on the account, integration and organization policies. Review the permissions actually requested before use.
Configuration covered: Qwen Code as a local command-line tool with a model access of your choice
Sources reviewed:
Show all 3 points
Documented
Account used
You sign in through /auth: with QwenCloud or Alibaba ModelStudio, a third-party provider or your own compatible endpoint. The free Qwen OAuth tier was discontinued on 15 April 2026. API keys are read from environment variables or a .env file; settings.json is a lower-priority fallback.
EvidenceDocumented with conditions
Admin approval
Control lies in settings files. Operators can set, among other things, the approval modes offered by the Web Shell and a Linux sandbox policy in the system scope. According to the page, the mode selection is presentation only and does not enforce backend permissions. The page does not describe an organisation approval.
EvidenceDocumented with conditions
Data access & team sharing
In YOLO mode the AI can execute any command with your terminal permissions. Session backups for /rewind live under ~/.qwen/file-history. The reviewed pages do not describe team sharing; the experimental qwen serve daemon lets several clients share sessions and by default listens on loopback without bearer authentication.
Evidence · Evidence · Evidence
Memory and conversation continuity
5 of 6 criteria documented
Instruction files plus automatic Markdown memories and optional structured recall.
Qwen Code documentation dated 2026-10-22
Sources accessed:
Documentation by criterion
StorageDocumented
Storage and scope
QWEN.md contains instructions; auto-memory is Markdown under ~/.qwen/projects/<project>/memory/.
SelectionDocumented
Selection and timing
Auto-memory saves useful preferences, feedback and context after conversations; it is enabled by default. Periodic maintenance has a separate toggle.
Edit & deleteDocumented
Inspect, edit and delete
/memory opens management and files; /remember, /forget and /dream control saving, forgetting and maintenance. Markdown can be edited directly.
Access boundariesDocumented
Agent and channel boundaries
Branches in one checkout share memory; linked worktrees have their own memory. Optional team memory is shared through Git.
RetrievalDocumented
Retrieval
The default is a MEMORY.md index. Optional structured recall injects relevant subtrees and adds search_memory for full entries.
New conversationsPartly documented
Session reset and compaction
New sessions begin with fresh context and reload durable knowledge. /dream maintains memories; this memory page does not explain conversation-window compaction.
Cost and effort
Prices as of 19 October 2026
Prices and included usage
Pricing checked
Apache-2.0 software; model and operating costs are separate.
- Billing depends on plan, usage and setup.
Setup
Install and connect an API provider or local model server.
Operation & oversight
Based on public provider documentation. Assessments are editorial inferences from the stated properties. Actual time requirements and reliability have not been measured.
Deployment covered: Local operation in the terminal and in headless mode (qwen -p)
Sources reviewed:
Editorial assessment
Operation and control are with you: you set approval mode, sandbox, budgets and scheduling yourself. Scheduled tasks do not survive ending the session; unattended runs need budgets and an external trigger such as cron or CI.
Show all 6 points
Documented with conditions
Files and working state
Saving chat history to disk is on by default. Session backups used by /rewind are retained for 30 days by default. Turning recording off also stops --continue and --resume from working.
EvidenceDocumented with conditions
Sign-ins
The page gives no expiry for API keys. Previously cached Qwen OAuth tokens may keep working briefly, but new requests are rejected.
EvidenceDocumented with conditions
Starting further runs
Runs start interactively or by headless call (qwen -p) from scripts and automation; the qwen serve daemon is labelled experimental. Scheduled tasks (/loop, cron tools) are on by default but live only in the running process and vanish on exit; recurring tasks expire after 7 days by default.
Evidence · Evidence · EvidenceDocumented with conditions
Human intervention
There are five modes: Plan, Ask Permissions, Auto-Edit, Auto and YOLO. Which mode is the default is inconsistent within the page. In YOLO mode the AI can execute any command with your terminal permissions.
EvidenceDocumented
Detecting errors
Headless mode produces text, JSON or stream-JSON with usage data. Budgets for wall time, tool calls and turns end with exit code 53 or 55, an interrupt with 130. An exit code for ordinary failures is not documented there.
EvidenceDocumented with conditions
Resuming after failures
With --continue or --resume you resume sessions with their history. /rewind rolls the conversation back to an earlier turn; /restore reverts project files from a checkpoint, but checkpointing is off by default. For unattended runs the variable QWEN_CODE_UNATTENDED_RETRY retries requests with delays capped at 5 minutes.
Evidence · Evidence · Evidence
Evidence and open questions
1 source · 1 open point
Open points
The reviewed sources contain no reliable information on these points. A gap says nothing about the product.
Requirements and limits: not yet confirmed
Sources and evidence
Provider sources reviewed; examples are editorial suggestions. No hands-on test.
Editorial team: Michael Meierhoff · sources reviewed on . Provider information is not independent evidence of performance.
Provider
Product and provider are documented through the linked original sources.
- Operator
- Alibaba / Qwen
- Organization
- Provider
- Website
- github.com
Changes to this profile 1
Added Qwen Code with capabilities, access, pricing and sources.
These notes describe editorial changes. Source and pricing review dates appear in their respective sections.