Qwen Code

An open-source coding agent for terminal and editor work, with optional headless runs and persistent Goals.

No product visual on file

Choose it when

You want repo work in a terminal or editor, and can explicitly choose the approval mode and provider before it starts writing.

Skip it when

You need an always-on service with verified failure ownership, or a guarantee that every edit and shell command waits for manual approval.

01 / Classification

Autonomy evidence

This is our classification, not a vendor badge. The record says what the product normally does. The gaps stay visible.

How we classify it

Tool use + multi-step plans, still human-gated per session.

Runs unattended

Yes — that capability is documented in this brief.

Trigger

A user starts a terminal/editor session or supplies a headless prompt via -p or stdin. Optional /goal work starts from an explicit objective or resume command.

State

Project-scoped JSONL histories restore conversation, tool outputs and compaction checkpoints. Goal state persists with the session; continuation requires chatRecording.

Tools

Documented write_file/edit tools change files; run_shell_command executes commands. Approval mode and tool restrictions govern these capabilities, not an assumed OS sandbox.

Human control

Plan is read-only; Ask manually gates edits/shell; Auto-Edit permits edits but gates shell; Auto uses a classifier; YOLO auto-approves both. Default wording conflicts, so select a mode. Goal work can be paused/resumed; explicit wall/tool budgets apply to single-shot runs, not every daemon/ACP path. No desk validation of classifier safety or failure recovery.

02 / Operations

Control reality

Useful agents eventually touch something real. This is the operational picture we have, not a promise hidden in a sales page.

Where it runs

Terminal/editor client with configurable third-party or local model endpoints; local client execution isn't a promise of local-only inference.

Data boundary

Authentication and chosen provider determine applicable privacy terms. Local execution doesn't mean code stays local; retention and provider-specific handling remain unverified here.

Identity & permissions

Uses the selected model-service authentication, including provider API keys and Vertex AI. Qwen OAuth's free tier is discontinued; no paid signup or entitlement was tested.

What it can write

Create and overwrite project files · Edit existing file contents · Execute shell commands

Surfaces

Terminal · Editor integrations · Optional headless scripts/CI

Export / lock-in

Headless text/JSON output and project-scoped JSONL histories provide output/history access in that scope, not verified cross-product session portability.

03 / Receipts

Evidence ledger

Here is what this brief is built on. “Documented” means we found a source. “Tested” means this desk actually ran it. Those are different things.

Evidence status

documented

Source set

11 primary sources

Evidence review

2026-10-31

Sources on file

Corrections

2026-10-31 · Added CLI-scoped release metadata: tagged v0.26.4 published 2026-10-29T14:29:11Z and npm latest @qwen-code/qwen-code 0.26.4 in the retained captures. This is not an installed-version observation or fresh validation of the previously cited capability docs. No installation or runtime test.

04 / Change log

Related coverage

The brief tells you what the product is. Payloads track the changes that may alter the decision.

No linked change coverage on file yet.