What is Qwen Gate?

Qwen Gate is a self-hosted API gateway that provides OpenAI-compatible access to Qwen large language models hosted on chat.qwen.ai. It uses browser automation to authenticate with Qwen accounts and then routes API requests through a browserless fetch layer. The gateway supports streaming SSE responses, OpenAI-style function calling parsed from model output, multi-account round-robin rotation with automatic failover, session pooling for efficiency, and a web dashboard for monitoring and account management. It runs on Bun with TypeScript, requires no build step, and can be installed via Docker, a one-command shell script, or manually on Linux, macOS, and Windows. The project is MIT-licensed and intended for educational use; users must follow chat.qwen.ai's terms of service.

Should you self-host Qwen Gate?

Qwen Gate scores 46/100 on selfhost.directory (how scoring works) and ranks #70 of 125 in AI. It has 206 GitHub stars.

It shipped 10 stable releases in the past year. The latest, v0.8.0, came out on 25 June 2026. New versions land roughly every 2 days.

Setup is moderate: expect some configuration beyond a single command. It is MIT-licensed and written mostly in Typescript.

Its closest alternative on selfhost.directory is Text Generation WebUI (76/100, 47 726 stars). It scores higher overall, so compare the two before you pick. See Qwen Gate vs Text Generation WebUI side by side.

Features

Cluster mode
Content filter pipeline
File upload for large context
Multi-account rotation3+ accounts recommended
No Build Step
OpenAI Compatible API
Session pooling
Streaming (SSE)
Tool CallingQwen doesn't natively support tool calling; gateway parses JSON text from model output
Web Dashboard

How to configure Qwen Gate

API_KEYBearer token for API auth (empty = no auth)
Bearer token for API auth (empty = no auth)
AUTH_REFRESH_BEFORE_MS300000Auth token refresh before expiry in milliseconds
Auth token refresh before expiry in milliseconds
AUTH_TOKEN_MAX_AGE_MS28800000Auth token max age in milliseconds
Auth token max age in milliseconds
CLAUDE_CODE_PROXYfalseEnable Claude Code proxy
Enable Claude Code proxy
values: true, false
CLEAN_OUTPUTtrueStrip internal artifacts from responses
Strip internal artifacts from responses
values: true, false
CUSTOM_INSTRUCTIONCustom system instruction
Custom system instruction
DARK_MODEfalseEnable dark mode in dashboard
Enable dark mode in dashboard
values: true, false
DELETE_SESSIONtrueDelete session on cleanup
Delete session on cleanup
values: true, false
HOSTBind address
Bind address
MAX_LOGS50Maximum logs to keep in memory
Maximum logs to keep in memory
MAX_TOOL_CALLS_PER_RESPONSE3Maximum tool calls per response
Maximum tool calls per response
MODELS_CACHE_TTL_MS3600000Models cache TTL in milliseconds
Models cache TTL in milliseconds
OPEN_DASHBOARD_ON_STARTfalseAuto-open dashboard in browser on start
Auto-open dashboard in browser on start
values: true, false
PORT26405Server port
Server port
QWEN_FETCH_TIMEOUT_MS30000Qwen fetch timeout in milliseconds
Qwen fetch timeout in milliseconds
RATE_LIMIT_COOLDOWN_MS120000Cooldown after rate limit in milliseconds
Cooldown after rate limit in milliseconds
RETRY_BACKOFF_MULTIPLIER2Retry backoff multiplier
Retry backoff multiplier
RETRY_BASE_DELAY_MS1000Base retry delay in milliseconds
Base retry delay in milliseconds
RETRY_ENABLEDtrueEnable retries
Enable retries
values: true, false
RETRY_MAX_ATTEMPTS3Max retry attempts
Max retry attempts
RETRY_MAX_DELAY_MS30000Max retry delay in milliseconds
Max retry delay in milliseconds
SAVE_REQUEST_LOGSfalseSave per-request logs to disk
Save per-request logs to disk
values: true, false
STREAM_IDLE_TIMEOUT_MS60000Stream idle timeout in milliseconds
Stream idle timeout in milliseconds
STREAMING_MODEautoStreaming mode
Streaming mode
values: auto, on, off
TOOL_CALLINGtrueEnable tool call parsing
Enable tool call parsing
values: true, false
USE_CUSTOM_INSTRUCTIONfalseUse custom instruction
Use custom instruction
values: true, false