zhouyoukang1234-spec/windsurf-assistant

Analysis updated 2026-06-24

★ 85JavaScriptAudience · developerComplexity · 4/5LicenseSetup · hard

Mindmap

mindmap
  root((windsurf-assistant))
    Inputs
      Windsurf accounts
      GitHub PAT
      System prompts
    Outputs
      OpenAI v1 endpoint
      Rotated sessions
      Rewritten prompts
    Use Cases
      Use Windsurf models in any OpenAI client
      Rotate accounts to avoid quota limits
      Inject custom system prompts into Cascade
    Tech Stack
      JavaScript
      Node.js
      Cloudflare
      GitHub Pages

mindmap root((windsurf-assistant)) Inputs Windsurf accounts GitHub PAT System prompts Outputs OpenAI v1 endpoint Rotated sessions Rewritten prompts Use Cases Use Windsurf models in any OpenAI client Rotate accounts to avoid quota limits Inject custom system prompts into Cascade Tech Stack JavaScript Node.js Cloudflare GitHub Pages

Click or tap to explore — scroll the page freely

What do people build with it?

USE CASE 1

Expose Windsurf models behind an OpenAI-compatible /v1 endpoint so LobeChat, Cursor, or Aider can call them.

USE CASE 2

Rotate between several Windsurf accounts you own with a 60-second lock so two sessions never collide.

USE CASE 3

Rewrite the Windsurf Cascade system prompt to a custom one while keeping @-tool integrations intact.

USE CASE 4

Run a fleet of Windsurf account instances on one VM behind a Cloudflare tunnel.

What is it built with?

JavaScriptNode.jsCloudflareGitHub-Pages

How does it compare?

	zhouyoukang1234-spec/windsurf-assistant	energypantry/agent-browser-runtime	joeseesun/qiaomu-userscripts
Stars	85	86	77
Language	JavaScript	JavaScript	JavaScript
Setup difficulty	hard	hard	easy
Complexity	4/5	4/5	2/5
Audience	developer	developer	general

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · hard Time to first run · 1h+

Needs your own VM, a fine-grained GitHub PAT, GitHub Pages fork, and a private Gist, pieces install separately and the project recently went through a self-audit after an abuse flag.

MIT license, very permissive, use freely with attribution.

In plain English

Windsurf Assistant is a self-hosted toolkit, written mostly in JavaScript with no npm dependencies, that sits around the Windsurf IDE and the AI models it talks to in the cloud. The README is partly in Chinese and styles itself around a Daoist motif, but underneath it bundles three orthogonal pieces that can be installed alone or together. The May 2026 compliance note at the top of the README explains that high-frequency cron schedules were turned off and the project went through a self-audit after a false-positive abuse flag. The first piece, called dao-core or the reverse API, runs on a machine you own and exposes Windsurf's models behind an OpenAI-compatible /v1 endpoint with streaming responses. Once it is running, any OpenAI client like LobeChat, OpenWebUI, NextChat, Cherry Studio, the openai SDK, Continue.dev, Aider, or Cursor's OpenAI override can be pointed at it and use the Windsurf models as if they were OpenAI's. The second piece, WAM, is a Windsurf IDE extension that rotates between multiple Windsurf accounts you own, using a 60-second lock so two sessions do not grab the same account, and switching when one runs out of quota. The third piece, dao-proxy-min, is a small proxy that sits between the Windsurf Cascade panel and the cloud and rewrites the system prompt to one of your own choosing while keeping the @-tool integrations intact. The README's example system prompt is the Mawangdui silk-text version of the Tao Te Ching. Distribution is unusual: there is a public entry page on GitHub Pages, you sign in with a fine-grained GitHub personal access token kept only in your browser, and the page walks you through forking the repo into your own account and turning on Pages. Your settings, virtual machine URL, accounts, system prompt presets, and chat history live in a private Gist on your own GitHub, not on any shared server. A daemon is provisioned on a VM you own with a one-line curl command that downloads a bootstrap script, and a fleet variant runs several account instances on one VM behind a Cloudflare tunnel. The code is released under MIT.

Copy-paste prompts

Prompt 1

Set up dao-core on a VM I own, then point a LobeChat client at the /v1 endpoint and chat using my Windsurf models.

Prompt 2

Install the WAM Windsurf extension and configure it to rotate between three Windsurf accounts with the 60-second lock.

Prompt 3

Run dao-proxy-min and rewrite the Cascade system prompt to my own text while preserving the @-tool integrations.

Prompt 4

Fork the windsurf-assistant repo into my account, turn on GitHub Pages, and bootstrap the daemon on a VM with the one-line curl command.

Prompt 5

Walk me through storing my windsurf-assistant settings in a private GitHub Gist using a fine-grained personal access token.

Frequently asked questions

What is windsurf-assistant?

Self-hosted toolkit that exposes Windsurf IDE models behind an OpenAI-compatible endpoint, rotates between Windsurf accounts, and rewrites Cascade system prompts.

What language is windsurf-assistant written in?

Mainly JavaScript. The stack also includes JavaScript, Node.js, Cloudflare.

What license does windsurf-assistant use?

MIT license, very permissive, use freely with attribution.

How hard is windsurf-assistant to set up?

Setup difficulty is rated hard, with roughly 1h+ to a first successful run.

Who is windsurf-assistant for?

Mainly developer.

Open on GitHub → Explain another repo

This repo across BitVibe Labs

Verify against the repo before relying on details.