HagiCode Product Overview

HagiCode keeps AI from losing context when it changes your code: understand the repository first, set the scope, then execute with a traceable trail. It is an AI workspace for project development that connects the Agent CLI and models you already use, so you do not need to switch editors. Rich, convenient UI interactions make planning, agent execution, and project work approachable and easy to navigate, creating an engaging development experience.

Install HagiCodeWatch the demo video

• Supports 11 Agent CLIs (Claude Code, Codex, GitHub Copilot, and more)
• Works on Windows, macOS, and Linux
• Core features are free

Core features at a glance

1. 01Understand first, then actConversation sessions are read-only by default; complex changes go through a proposal first.
2. 02One scope across repositoriesMonoSpecs treats several repositories as one project, with specs kept in the main repository.
3. 03Keep your Agent CLI11 Agent CLIs supported, health visible at a glance, no editor switch.
4. 04SkillsSearch, then check the source and command before installing and decide which sources to trust.
5. 05Vault and Code ServerBring material from outside the repository into a session and edit it in the browser.
6. 06Super CommitGrouped Conventional Commits, including across several repositories.
7. 07Visible progressRich, convenient UI interactions make project work easier to follow, with progress feedback as a supporting detail.

How a development task works

For example, suppose you need to change a feature spanning several repositories:

1Understand first

A conversation session where the agent searches the repositories with a tool call and explains where invoices are rendered

Analyze the existing code in a conversation session. Conversation sessions are read-only by default; switch to edit mode when you need to change files.

2Define the scope

New proposal drawer showing repository selection and a proposal preview

For a complex change, select the project and repositories in a proposal session. Outline the goals, tasks, and verification steps before proceeding with execution.

3Track the outcome

Session detail after a proposal workflow has finished executing

Review sessions and execution status in the workspace. When you need to prepare a commit, use Super Commit.

4Keep the context

Session board and repository status of a MonoSpecs project

Revisit proposals and commit history. For cross-repository work, use MonoSpecs to manage repository relationships and change scope.

You do not need a proposal for every task. Start with a conversation session when you only want to explore the code or handle a simple issue. HagiCode provides a traceable path when a task needs more context.

Features in detail

Understand first, then act

HagiCode separates “reading the code” from “changing the code” so you can choose by the size of the task instead of letting AI edit everything directly.

• Conversation session: Read-only by default. Use it to explore a repository, investigate an issue, or ask where some logic lives. Switch to edit mode only when you need to change files.
• Proposal session: For complex changes. Choose the project and repository scope, outline the goals, tasks, and verification steps, then move on to execution once you are happy with the plan.
• Session board: Proposal & Review, Execution, and Archived lanes show where every session stands at a glance. You can switch between the Classic, Execution Flow Kanban, and Dashboard Kanban layouts.

Learn more: Create a conversation session · Create a proposal session · The session board

One scope across repositories

When a request touches the frontend, the backend, and the docs at once, MonoSpecs lets HagiCode understand them as a single project.

• Register sub-repository locations, Git URLs, and display names in .hagicode/monospecs.yaml.
• Change specs (OpenSpec) live in the main repository, so the code directories of the sub-repositories stay clean.
• When you create a proposal, choose which sub-repositories it involves. The project page shows the branch and uncommitted changes of each repository.

Learn more: MonoSpecs guide

Keep your Agent CLI

HagiCode does not replace your toolchain. It brings it into one workspace.

• An Agent is a CLI, a model, and optional personality text. Create several Agents, assign them to execution scenarios, and set up backup groups.
• Supports 11 Agent CLIs, including Claude Code, Codex, and GitHub Copilot. The workspace shows each CLI’s version, response time, and health.
• With OmniRoute, you can connect several OpenAI-compatible tools and models to one local gateway, while CLI and model connections stay managed separately.

Learn more: Agents workspace · AI Agent CLI installation and support

Skills

The Skills panel puts the skills already on your machine and the skills you can install from the gallery behind one entry.

• Local Skills: See what is installed on this machine and check for updates in batch.
• Gallery: Search by keyword and source, and see whether a skill is already installed.
• Trusted Providers: Before installing, confirm the source, trust status, and the command that will actually run, and decide which sources count as trusted.

Learn more: Skills guide

Vault and Code Server

• A Vault is a folder outside your project repositories that HagiCode may browse. Use it for reference documents, notes, and research output. Three types are supported: Folder, CodeRef, and Obsidian.
• When you create a session, choose which Vaults to include in its scope. Mark one as editable for AI to write to it; otherwise it is read-only reference.
• Code Server provides an editing entry point in the browser that you can open straight from a Vault or project.

Learn more: Vault guide

Super Commit

Click “Super Commit” in the Git tab and HagiCode reads your current changes and writes commit messages in Conventional Commits format.

• Related changes are grouped into several commits instead of one catch-all.
• In MonoSpecs projects, changes are committed to the right sub-repository.
• A Co-Authored-By tag is added automatically, and its name and email can be set per project or globally.

Learn more: Super Commit guide

Visible progress

Rich, convenient UI interactions make planning, agent execution, and project work approachable and easy to navigate, creating an engaging development experience. Achievements, daily reports, and Token metrics add supporting feedback for long-running work. To see how Agents are composed, assigned, and configured, read Agents and related configuration. These game elements support the workspace experience, and HagiCode remains usable as a regular development tool without them.

Learn more: Status bar and progress

Who it is for

• A good fit: Developers and teams who maintain several repositories and want complex changes planned before they are executed, and people who already use Agent CLIs such as Claude Code or Codex and want them in one workspace.
• Not essential: If you only need a few lines completed in your editor, your current tools are enough.

See Hagicode in real coding sessions

Start with the English YouTube overview embedded below, then compare two supporting Bilibili demos without losing the product story or the direct watch links.

Featured walkthroughYouTube

Hagicode Product Overview on YouTube

This is the primary English-language product introduction, so readers can understand Hagicode's workflow, product scope, and multi-agent angle before leaving the docs page.

Open on YouTube

Focused demoBilibili

AI Playing Games While Coding

A supporting Bilibili demo that shows the product feeling playful and alive during real coding sessions instead of behaving like a dry code generator.

Open on Bilibili

Focused demoBilibili

GPT Codex in Hagicode: Live Trial

A narrower validation run that helps readers judge how GPT Codex behaves inside the actual Hagicode product after the main YouTube introduction.

Open on Bilibili

Get started in three steps

1. Choose an installation methodFind the right option for your operating system and deployment environment.

2. Complete the setup wizardConfigure the basics after first launch and create at least one project.

3. Create a conversation sessionStart by exploring your project in read-only mode, then decide whether to edit or create a proposal.

Editions & Pricing

Core features are free. Desktop and Microsoft Store are two different ways to obtain HagiCode, not two paid tiers. The Microsoft Store channel also offers Turbo Engine, a separate one-time purchase that unlocks a higher proposal concurrency limit and enhancements such as documentation themes, custom attribution, branding, and custom avatar uploads. See the Turbo Engine add-on for the specific features it unlocks.

Your environment	Recommended option
Windows with access to Microsoft Store	Install from Microsoft Store
Manual Windows installation, or local use on macOS or Linux	Install Desktop

The Steam main application has been retired and is no longer an installation channel. See Steam distribution status.

Why make it a workspace?

I’m Kun Yu, the creator of HagiCode. A single question and answer can solve an immediate problem, but real development also means understanding the project, setting boundaries, carrying changes through, and reviewing the result. I wanted to bring these scattered steps together; gamified feedback makes progress in long-running work easier to see. You can use just some of these features. You do not have to learn every module before you start coding.