Hagicode Product Overview

In one sentence, Hagicode is an AI coding assistant built around real development workflows. It helps teams understand code, organize changes, execute tasks, and keep the reasoning behind the work.

This page does not try to list every module. It stays with three themes: Smart, Efficient, and Fun.

What is Hagicode?

Hagicode is not just a tool for generating a few lines of code. It is a workspace for the broader development loop:

• Understand codebases before changing them
• Turn ideas into structured changes with proposals, tasks, and validation
• Keep work traceable so decisions can be reviewed and reused later
• Connect to current supported Agent CLIs: Claude Code, Codex, GitHub Copilot, OpenCode, Hermes, QoderCLI, Kiro, Kimi, Gemini, DeepAgents, and Codebuddy

Hermes, Gemini, and DeepAgents are supported through manual local installation or authentication flows. The remaining CLIs expose one-click installation entry points or guided install flows. For setup details, use the AI Agent CLI Installation page. IFlowCli remains legacy compatibility only and is not part of the current active support range.

Smart: understand first, then act

The “Smart” part of Hagicode is less about flashy demos and more about making AI useful in real projects.

Align on goals, scope, and validation before implementation starts.

1. Proposal-driven instead of blindly editing

For larger changes, Hagicode can turn an idea into a proposal before implementation starts. That gives you a shared view of:

• the goal
• the scope
• the task breakdown
• the validation criteria

This makes AI feel more like a planning partner than an auto-editor.

2. Read-only / edit dual mode

Sometimes you want analysis, not changes. Hagicode separates those two cases clearly:

• Read-only mode for exploring unfamiliar code, tracing bugs, and understanding architecture
• Edit mode for implementing features, fixing defects, and refactoring with intent

That boundary makes it easier to trust the workflow.

3. Project context matters

Hagicode focuses on project-level understanding instead of isolated code snippets:

• it looks at repository structure and existing patterns
• it supports multi-turn conversation, task tracking, and tool usage
• it works better as a long-running engineering partner

4. Team knowledge stays visible

Proposal and archive flows help preserve the reason behind a change, not just the final diff. That matters when teams grow, hand work off, or revisit earlier decisions.

Efficient: turn common actions into one smooth flow

The “Efficient” part of Hagicode is about reducing context switching and helping work move in a natural sequence.

Understand first, then choose whether the AI should move into editing work.

The current product already combines session progress, project status, and repository context into one board.

1. One workflow from understanding to delivery

A common workflow looks like this:

1. import or create a project
2. inspect the code in a read-only session
3. break the change down in a proposal session
4. switch to edit mode to implement
5. generate a better commit message with AI

That is simpler than spreading the same work across disconnected tools.

2. Monospec for multi-repository work

If your work spans more than one repository, Monospec gives you a unified view:

• coordinate multiple repositories in one place
• keep conventions aligned across repos
• give AI a wider context for cross-repo analysis

It works well for microservices, multi-module products, and long-lived platform work.

3. AI Compose Commit for cleaner history

Good code changes still need clear commit history. Hagicode helps generate commit messages that are easier to read and easier to maintain:

• closer to Conventional Commits style
• clearer for reviewers and teammates
• more useful for release notes, rollback, and audits

4. Clear setup paths

Whether you prefer Desktop, Docker Compose, or a specific Agent CLI stack, the docs aim to keep the setup path explicit so you can start faster with less guesswork.

Who is Hagicode for?

Role	What you get
New engineers	Faster onboarding into unfamiliar codebases
Developers	Less switching between analysis, implementation, and commit cleanup
Technical leads	Better control over complex changes and decision traceability
Multi-repo teams	A more unified way to coordinate work across repositories

Start here

Quick start

Wizard Setup The current canonical entry: finish first-project onboarding in the wizard and bind it to a real repository.

Create a Conversation Session Use the default read-only flow to understand a project first, then switch to edit mode when execution is needed.

Create a Proposal Session Turn vague ideas into an executable plan for larger or collaborative work.

Installation options

Desktop A guided desktop flow for users who want to get started quickly.

Docker Compose A good fit for isolated environments, shared setups, or team deployment.

Explore key capabilities

Monospec See how Hagicode approaches multi-repository collaboration.

AI Compose Commit Learn how AI can help produce cleaner, more useful commit messages.

Skills See local skills, recommendations, and the gallery for expanding the workspace over time.

Real Product Walkthroughs

See Hagicode in real coding sessions

Start with the core daily workflow, then jump into a playful AI side quest and a GPT Codex validation run inside Hagicode.

Featured walkthrough

Daily Hagi Half-Hour: AI Multi-Task Programming Practice

A guided overview of Hagicode's everyday workflow, agent orchestration, and the core experience new visitors should understand first.

Open on Bilibili

Focused demo

AI Playing Games While Coding

A playful clip that shows how the assistant can stay lively and surprising during real coding sessions instead of feeling like a dry code generator.

Open on Bilibili

Focused demo

GPT Codex in Hagicode: Live Trial

A focused validation run of GPT Codex inside Hagicode so prospective users can judge how the model performs in the actual product.

Open on Bilibili

---

For deeper reading, start with quick start, then open the guide for the capability you actually plan to use next.

Live broadcast notices are optional. The default docs build keeps this page quiet; set VITE_FEATURE_LIVE_BROADCAST_ENABLED=true when you want to surface the live entry here again.