Use Claude Code Agent Team mode to give one person a full AI development team. Works for all software development projects (iOS / Android / macOS / Windows / Web / cross-platform).
In Greek mythology, the hero Cadmus slew a dragon and sowed its teeth into the earth. A host of fully armed warriors sprang from the ground — they were called Spartoi (Σπαρτοί), meaning "the sown ones."
This is the same story as iSparto's workflow: you sow your product requirements into /init-project, and an entire Agent Team assembles itself — Lead breaks down tasks, Developer writes code, Codex reviews and fixes, Doc Engineer keeps documentation in sync — a complete development team grown from a single seed.
The i was moved from the end of Spartoi to the front. Lowercase i = I = me, one person.
iSparto = I + Sparto = one-person army.
flowchart TB
User["User<br/><br/>Product direction<br/>Requirements definition<br/>Wave boundary acceptance"]
subgraph Lead["Team Lead — Claude main session"]
direction LR
LeadTasks["Break down tasks<br/>Interface contracts<br/>Coordinate full workflow<br/>Merge code"]
Doc["Doc Engineer<br/>sub-agent<br/><br/>After Wave completion<br/>Documentation audit"]
end
Dev["Claude Developer<br/>teammate<br/><br/>Write code<br/>Write unit tests<br/>Review Codex fixes"]
Codex["Codex 5.3 Reviewer<br/>MCP call<br/><br/>Code review<br/>Fix issues directly<br/>QA smoke testing"]
User --> Lead
Lead --> Dev
Lead --> Codex
- Lead / Developer / Doc Engineer: Claude Opus 4.6 + max effort
- Codex Reviewer: Codex 5.3 (via MCP, using $20 ChatGPT subscription, xhigh reasoning + fast mode)
Existing AI coding tools (Cursor, Windsurf, Copilot, Claude Code single session) all follow the same pattern — you go back and forth with a single Agent. The Agent has no team, no division of labor; everything depends on you and it trading messages back and forth.
iSparto turns a single Agent into a team with clear roles: Lead breaks down tasks, Developers write code in parallel, Codex cross-reviews, and Doc Engineer keeps documentation in sync. Instead of directing an Agent line by line, you confirm the direction and accept the results.
| Single-Agent Tools | iSparto | |
|---|---|---|
| Collaboration mode | You go back and forth with a single Agent | You work with the Lead; Lead coordinates the whole team |
| AI organization | Single Agent, no division of labor | Team-based (Lead + Developer + Reviewer + Doc Engineer) |
| Parallelism | None — single-threaded conversation | Multiple Developers in parallel within a Wave, visible via tmux split panes |
| Code review | Agent reviews its own code (same source) | Codex reviews Claude (different source), covering each model's blind spots |
| Cross-session state | Lost — must re-explain context every time | Driven by plan.md; /start-working auto-restores state |
| Documentation sync | Manual maintenance | Doc Engineer auto-audits every Wave |
In short: other tools have you directing one Agent. iSparto has you directing an entire team.
| Item | Requirement | Notes |
|---|---|---|
| Claude Max subscription | $100/month | Claude Code + Agent Team mode |
| ChatGPT subscription | $20/month | Codex CLI (code review + QA) |
| Node.js | 18+ | Runs Claude Code, Codex CLI, and MCP Server |
| Git | Any version | Version control |
| Terminal | iTerm2 (macOS) | Agent Team tmux mode relies on iTerm2's built-in tmux integration; no separate tmux installation needed |
Total cost: $120/month — two top-tier models (Claude Opus + Codex), no additional API fees.
curl -fsSL https://raw.githubusercontent.com/BinaryHB0916/iSparto/main/install.sh | bashOne command handles everything: downloads iSparto to ~/.isparto, checks/installs Claude Code and Codex CLI, logs into Codex, copies commands and templates to ~/.claude/, and registers the global MCP Server. Your existing ~/.claude/settings.json is never modified.
Alternative: manual clone
git clone https://github.com/BinaryHB0916/iSparto.git
cd iSparto && ./install.shmkdir my-app && cd my-app
claude --effort max
/env-nogo # optional — confirm environment readiness
/init-project I want to build an xxx # generates CLAUDE.md + docs/, Codex architecture pre-reviewcd existing-project/
claude --effort max
/migrate # scans project, proposes migration plan, preserves all existing content/start-working
→ Lead reads plan.md, reports current status and TODOs
→ You confirm "go ahead"
↓
Lead's team runs on its own (you don't need to watch)
→ Break down tasks → Developer writes code → Codex reviews → Developer reviews fixes
→ Codex QA → Doc Engineer documentation audit → Lead merges code
↓
Occasionally Lead comes to you (escalate decisions / confirm commits)
↓
/end-working
→ Sync documentation → Update plan.md → commit → push
/plan I want to add an xxx feature
→ Lead first reviews the product direction, produces a proposal
→ After you confirm the proposal, Lead writes it into plan.md and begins work
One-time setup (./install.sh handles this automatically):
- Claude Max + ChatGPT subscriptions active
- Terminal is iTerm2 (macOS, required for Agent Team split panes)
-
./install.shcompleted (Claude Code, Codex CLI, config files, MCP) - Multi-device sync configured (if using multiple computers, see configuration.md)
Each new project (/init-project handles this automatically):
- Launch with
claude --effort max -
/env-nogocheck passed (optional) -
/init-projecthas generated CLAUDE.md + docs/ - Project-level
.claude/settings.jsonconfigured with platform-specific plugins (e.g., swift-lsp for iOS, optional)
iSparto/
├── README.md ← The document you are reading now
├── settings.json ← Reference template for project-level .claude/settings.json
├── CLAUDE-TEMPLATE.md ← Template for generating new project CLAUDE.md
├── LICENSE
├── .gitignore
├── install.sh ← One-click install script
├── commands/
│ ├── start-working.md ← Start working command
│ ├── end-working.md ← End working command
│ ├── plan.md ← Planning command
│ ├── init-project.md ← Initialize project command
│ ├── env-nogo.md ← Environment readiness check
│ └── migrate.md ← Migrate existing project to iSparto
├── templates/
│ ├── product-spec-template.md
│ ├── tech-spec-template.md
│ ├── design-spec-template.md
│ └── plan-template.md
└── docs/
├── concepts.md ← Core concepts (decoupling, Wave, file ownership) ⭐ Recommended reading
├── user-guide.md ← User interaction guide (6 commands + 3 notifications) ⭐ Recommended reading
├── roles.md ← Role definitions + Codex prompt templates
├── workflow.md ← Full development workflow + branching strategy + Codex integration
├── configuration.md ← Global configuration + adaptation guide + multi-device sync
├── troubleshooting.md ← Common troubleshooting
└── design-decisions.md ← Design decision records
