HagiCode turns ideas into shipped work with OpenSpec workflows, multi-agent parallel execution, and Hero Dungeon battle reports that keep AI coding visible, structured, and fun.
🌐 Visit Website · 🖥️ Desktop · 🐳 Container · 📚 Product Overview
- Smart - OpenSpec gives teams a structured path from idea, proposal, and design to implementation, testing, and archive.
- Efficient - Claude Code, Codex, and other CLI agents can drive multiple agents and instances in parallel, so proposals, fixes, and reviews keep moving together.
- Interesting - Hero Dungeon rosters, battle reports, and visual workspaces make daily AI coding feel more like a coordinated adventure than a pile of disconnected chats.
The gallery below uses site-owned assets copied into repos/site, so GitHub can render current product evidence without deep-linking into the docs repository.
Instead of relying on one hero screenshot or old metrics cards, this tour walks through the active proposal flow, quick execution mode, skill safety surfaces, Hero identity, and desktop controls.
Proposal stages at a glance
UI screenshot. Track proposal work from optimization through archive in one nine-stage timeline.
Inline review and annotations
UI screenshot. Review files with inline annotations before sending focused feedback back to AI.
Execution result with code diff
UI screenshot. Inspect the applied diff immediately after plan execution finishes.
Normal session execution result
UI screenshot. Use quick chat sessions for focused tasks when you do not need a full proposal flow.
Skill trust and managed command
UI screenshot. Verify the install source, trust status, and exact managed command for every skill.
Hero highlight and personality layer
Product visual. Hero Dungeon personas and battle-report theming give long-running AI sessions a memorable identity.
Desktop start service control
UI screenshot. Start or monitor the local HagiCode service from the desktop dashboard.
Desktop version management
UI screenshot. Switch installed packages, dependencies, and release tracks from the desktop version center.
- Proposal-driven by default - OpenSpec keeps scope, tasks, and archive history connected so AI work stays reviewable.
- Built for parallel throughput - the current product showcase highlights real-time token tracking, multi-agent workspaces, and execution views that are designed for multiple streams at once.
- Memorable to use - Hero Dungeon and Hero Battle report views add narrative, feedback, and momentum to team workflows.
Track how the HagiCode-org/site repository has grown over time:
Start with the official product entry points:
- hagicode.com for the full homepage experience
- Desktop to install the local app
- Container for the self-hosted deployment path
- Product Overview for the documentation entry
Run the site locally from repos/site:
npm install
npm run dev
npm run build
npm run previewThe dev server runs on http://localhost:31264 by default. For contributor-oriented details, start with AGENTS.md and CLAUDE.md.
The homepage activity metrics module is now a runtime consumer of the canonical JSON hosted by the index site:
- Source:
https://index.hagicode.com/activity-metrics.json - Single source of truth:
repos/indexgenerates and publishes the only homepage metrics asset thatrepos/siteis allowed to consume - Ownership: the data is generated and published from
repos/index, not fromrepos/site - Failure mode: if the remote asset is unavailable, the homepage keeps the existing empty-state fallback instead of breaking the rest of the page
- Maintenance boundary: if homepage metrics look wrong or stale, investigate the index repository and index deployment first instead of re-adding a local refresh script or local JSON copy in this repository
This repository is released under the terms in LICENSE.
Ready to explore the full product story? Visit hagicode.com and choose the experience that fits your workflow.