By shakestzd
Unified observability and orchestration for AI coding agents - yoking agents together
Add a new feature to the Ijoka graph database.
Report a bug discovered during work.
Spawn a Haiku subagent to intelligently classify recent unlinked activity and associate it with features.
Mark the currently active feature as complete in the Ijoka graph database.
Session end command - ensures proper handoff for the next session.
Matches all tools
Hooks run on every tool call, not just specific ones
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Unified observability and orchestration for AI coding agents - yoking agents together.
Ijoka (Zulu for "yoke") provides a lightweight desktop application and Claude Code plugin that implements Anthropic's long-running agent pattern for coordinating work across multiple AI coding assistants.

feature_list.json and session transcripts┌─────────────────────────────────────────────────────────────┐
│ Ijoka Desktop (Tauri) │
│ - Kanban Board UI │
│ - Activity Timeline │
│ - Progress Stats │
└─────────────────────────────────────────────────────────────┘
▲
SQLite + File Watcher + HTTP Server
│
┌─────────────────────┼─────────────────────┐
│ │ │
┌───┴───┐ ┌────┴────┐ ┌────┴────┐
│ Claude │ │ Codex │ │ Gemini │
│ Code │ │ CLI │ │ CLI │
└────────┘ └─────────┘ └─────────┘
│ │ │
└─────────────────────┴─────────────────────┘
│
feature_list.json
# Clone the repository
git clone https://github.com/Shakes-tzd/ijoka.git
cd ijoka
# Install dependencies
pnpm install
# Run desktop app in development
pnpm dev
# Build for production
pnpm build
# From the repo root
pnpm plugin:install
# Or manually
cd packages/claude-plugin
claude /plugin install .
ijoka/
├── apps/
│ └── desktop/ # Tauri desktop application
│ ├── src-tauri/ # Rust backend
│ │ ├── src/
│ │ │ ├── main.rs
│ │ │ ├── db.rs # SQLite database
│ │ │ ├── watcher.rs # File watching
│ │ │ ├── server.rs # HTTP server for hooks
│ │ │ └── commands.rs # Tauri commands
│ │ └── Cargo.toml
│ └── src/ # Vue frontend
│ ├── App.vue
│ └── components/
├── packages/
│ └── claude-plugin/ # Claude Code plugin
│ ├── .claude-plugin/
│ │ └── plugin.json
│ ├── hooks/
│ │ ├── hooks.json
│ │ └── scripts/
│ ├── commands/
│ ├── agents/
│ └── skills/
├── shared/
│ └── types/ # Shared TypeScript types
└── docs/ # Documentation
Ijoka implements Anthropic's recommended pattern for multi-session development:
feature_list.jsonA persistent task queue that survives across sessions:
[
{
"category": "functional",
"description": "User authentication with OAuth",
"steps": ["Create auth route", "Implement OAuth flow", "Add session management"],
"passes": false
}
]
feature_list.json, picks ONE feature where passes: falsepasses: false → true"We use strongly-worded instructions like 'It is unacceptable to remove or edit tests.' After experimentation, we landed on JSON as the model is less likely to inappropriately change or overwrite JSON files compared to Markdown." — Anthropic
| Command | Description |
|---|---|
/init-project | Initialize feature_list.json in current project |
/feature-status | Show completion percentage and next tasks |
/next-feature | Pick and start the next incomplete feature |
Settings are stored in:
~/Library/Application Support/com.ijoka.app/%APPDATA%\com.ijoka.app\~/.config/com.ijoka.app/Database is stored at ~/.ijoka/ijoka.db
Configure watched projects in .claude/settings.json:
npx claudepluginhub shakestzd/ijoka --plugin ijokaPrecision-tuned context engineering for Claude Code. Optimize context flow with modular plans (95% fewer tokens), parallel workflows (81% cost reduction), and zero-transformation architecture. Quick research (/ctx:research: 3 parallel agents, 1-2 min). Contextune your workflows for peak performance!
Local-first observability and coordination platform for AI-assisted development. Go binary hooks for near-zero cold start.
Quarto-based academic manuscript pipeline for journal submission. Scaffolds projects, builds publication-ready tables, detects AI writing tells, and generates upload packages. Agents: ai-tell-fixer, manuscript-reviewer, table-builder. Skills: manuscript-setup, build-upload, journal-compliance.
Research and publication pipeline — 14 agents, 13 skills, 5 commands covering data extraction, analysis, visualization, writing, peer review, and journal submission. Integrates flowmpl (design system), quartopress (manuscript pipeline), prs-dataviz (medical figures), parameterized-manuscript pattern, and a full research toolkit.
CRISPI plan authoring and review in a local marimo notebook. /plan creates or reopens a structured YAML plan; human approvals persist to .moplan/moplan.db on every click.
Lanes skills and slash commands for Claude Code
Long Task Harness for AI agents - task/feature-driven development with external memory
Long-running agent harness with 5-layer memory architecture, GitHub integration, autonomous batch processing, Agent Teams with ATDD, 9 hooks (safety, quality gates, team coordination), and 6 Agent Skills
Cotask — task management with TASKS.md kanban dashboard
Persistent memory, shared standards, and structured workflows for AI coding agents. Detects project setup and injects agent context automatically.
Task distribution, agent coordination, progress monitoring - executes plans via subagents. Requires AI Maestro for inter-agent messaging.