Download FylloCode 0.15.5 til Windows

In this new era, the work that really needs our attention is no longer repetitive coding, but higher-level business understanding and architecture design. The future workflow I have in mind is: understand the business → think through the architecture → write a Proposal → review the Proposal → let the Agent implement it automatically, then move on to the next Proposal. That thinking led me to build FylloCode — a step from Vibe Coding toward Agentic Coding.

How FylloCode Solves It FylloCode enforces a structured workflow that separates thinking from doing: ┌──────────┐ ┌────────────────┐ ┌───────────┐ ┌───────────┐ │ Task │ ──→ │ Chat/Proposal │ ──→ │ Apply │ ──→ │ Archive │ │ │ │ │ │ │ │ │ │ What to │ │ Agent explores │ │ Agent │ │ Specs │ │ work on │ │ codebase and │ │ implements│ │ updated, │ │ │ │ writes a plan │ │ the plan │ │ change │ │ │ │ │ │ │ │ recorded │ │ │ │ ➜ YOU REVIEW │ │ │ │ │ └──────────┘ └────────────────┘ └───────────┘ └───────────┘ Task — Pick a task from your local list or synced from platforms like Yunxiao. One click sends it to Chat. Chat / Proposal — The Agent explores your codebase, asks clarifying questions, and produces a structured proposal (what to change, which files, acceptance criteria). The Agent is prohibited from writing code in this stage. You review, edit, and approve the proposal before anything else happens. Apply — A fresh Agent session implements the approved proposal task-by-task. It reads only the proposal artifacts — not the Chat history — so every decision must be captured in writing, not left in context-window memory. Archive — Specs are updated, the change is recorded with full traceability. Your project's knowledge base grows with every shipped change. Why Separating Thinking from Execution Matters Most Coding Agents understand a task and write code in the same session. Any misunderstanding becomes code directly, and code is expensive to review. FylloCode physically separates understanding from execution. Misunderstandings can only become proposal text — and text is cheap to review. A 2-minute proposal review catches problems that would take 20 minutes to find in a diff.

Features

Agent Protocol (ACP) FylloCode connects to any Coding Agent through ACP (Agent Client Protocol). Claude Code, Codex, or any ACP-compatible agent — one protocol, one interface, without juggling a pile of terminal windows. There are currently 35 agents available in the ACP Registry, with support for parallel chats and parallel Proposal Apply runs.

Udgiverens beskrivelse (engelsk)

Projektbeskrivelse på GitHub: “FylloCode 是一个开源桌面应用，把 Claude Code、Codex 等本地 Coding Agent 接进项目工作流。派任务、审方案、追溯每一次改动，积累下来的规范和结论，换一个 Agent 也能继承。”

Fordele / Ulemper

Kort sagt: FylloCode er et værktøj med åben kildekode i kategorien AI-assistenter og LLM til pc’er med 64-bit. Administratorer kan udrulle det uovervåget.

Er det sikkert at downloade FylloCode?

• Filen leveres af github.com, udgiverens eget domæne
• Downloadlinket bruger HTTPS
• Der er offentliggjort en SHA256-hash, så du kan kontrollere filen: CB4C8074DE71A6D651D169A3AD097496A5CE777E61ED0F8DAC735D5A95F318B6 · Åbn VirusTotal-rapporten for denne hash
• Linket er ikke kontrolleret endnu
• Kildekoden er offentlig
• ExeAtlas hoster, indpakker eller ændrer ikke installationsfiler

Relateret vejledning: Sådan tjekker du en fils SHA256-hash i Windows

Relateret vejledning: winget: installer og opdater programmer i Windows

Information

Udgiver

Fio

Kategori

AI-assistenter og LLM

Licens

AGPL-3.0

Pris

Gratis

Kildekode

github.com/Fioooooooo/FylloCode

Programmeringssprog

TypeScript

Kræver

Windows 11 / 10

Arkitektur

64-bit

Installationsfil

EXE (NSIS)

Version

0.15.5

Udgivet

24. august 2026

I kataloget siden

1. juni 2026

Registrerede versioner

21

GitHub-stjerner

28

GitHub-forks

4

Websted

github.com

winget-id

Fioooooooo.FylloCode

Bedømmelser og anmeldelser

Ingen anmeldelser endnu. Vær den første til at bedømme programmet.

• 
  Claude Code2.1.302AI-assistenter og LLM · Anthropic PBC
  Proprietær67,1 t ↓
• 
  Codex CLI0.161.0AI-assistenter og LLM · OpenAI, Inc.
  Open source16,6 t ↓
• 
  oh-my-pi18.12.6AI-assistenter og LLM · Can Bölük
  Open source654 ↓
• 
  ChatGPT1.1.0AI-assistenter og LLM · lencx
  Open source17,1 t ↓
• 
  GPT4All3.10.0AI-assistenter og LLM · Nomic
  Open source8,7 t ↓
• 
  Jan0.8.7AI-assistenter og LLM · Jan
  Open source3,3 t ↓
• 
  OpenWork0.18.62AI-assistenter og LLM · differentai
  Open source333 ↓
• 
  WSL Manager2.5.0.0AI-assistenter og LLM · Bostrot
  Open source14 t ↓

Denne side på andre sprog

• English
• Türkçe
• Deutsch
• Français
• Español
• Português (Brasil)
• Italiano
• Русский
• Polski
• Nederlands
• 日本語
• 简体中文
• 한국어
• Bahasa Indonesia
• Tiếng Việt
• العربية
• Українська
• Čeština
• Magyar
• Ελληνικά
• Svenska
• Suomi
• Norsk bokmål
• Română
• ไทย
• हिन्दी
• தமிழ்
• বাংলা
• עברית
• فارسی
• اردو
• Català
• Srpski
• Slovenščina
• Eesti
• Bahasa Melayu
• Oʻzbekcha
• Қазақ тілі
• Български
• Slovenčina
• Hrvatski
• Lietuvių
• Latviešu
• Shqip
• Azərbaycan dili
• ქართული
• Հայերեն
• Беларуская
• Íslenska
• Afrikaans
• Kiswahili
• አማርኛ
• मराठी
• తెలుగు
• ਪੰਜਾਬੀ
• नेपाली
• မြန်မာ
• پښتو
• 繁體中文