https://github.com/hlsitechio/agentic-swarm
🐝 Write your AI agents once, deploy anywhere. 60 specialist engineering agents in 10 teams — one command into Claude Code, VS Code, Codex, OpenCode, Cursor & Pi.
https://github.com/hlsitechio/agentic-swarm
agent-personas ai-agents ai-personalities anthropic avengers claude claude-code code-review coding-fun developer-tools entertainment fun game-of-thrones gordon-ramsay harry-potter lord-of-the-rings mcp personality star-wars vikings
Last synced: 3 days ago
JSON representation
🐝 Write your AI agents once, deploy anywhere. 60 specialist engineering agents in 10 teams — one command into Claude Code, VS Code, Codex, OpenCode, Cursor & Pi.
- Host: GitHub
- URL: https://github.com/hlsitechio/agentic-swarm
- Owner: hlsitechio
- Created: 2026-02-19T03:08:23.000Z (4 months ago)
- Default Branch: main
- Last Pushed: 2026-06-09T16:00:42.000Z (12 days ago)
- Last Synced: 2026-06-09T16:12:12.637Z (12 days ago)
- Topics: agent-personas, ai-agents, ai-personalities, anthropic, avengers, claude, claude-code, code-review, coding-fun, developer-tools, entertainment, fun, game-of-thrones, gordon-ramsay, harry-potter, lord-of-the-rings, mcp, personality, star-wars, vikings
- Language: Python
- Size: 211 KB
- Stars: 1
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# 🐝 Agentic Swarm
### Write your AI coding agents **once**. Deploy them into **any** tool in seconds.
Claude Code · VS Code · Codex · OpenCode · Cursor · Pi — same agents, one command.
[](https://code.claude.com)
[](https://code.visualstudio.com)
[](https://developers.openai.com/codex)
[](https://opencode.ai)
[](https://cursor.com)
[](https://pi.dev)
[]()
[]()
[]()
[]()
---
## 🎯 The problem
You tune a great set of agents — a sharp code reviewer, a security auditor, a database expert.
Then you try a new tool, and… you start over.
Every assistant stores agents differently: **Claude Code** wants `~/.claude/agents/*.md`, **VS Code**
wants `.github/agents/*.agent.md`, **Cursor** wants `.cursor/rules/*.mdc`, **Codex** wants
`~/.codex/prompts/`, **OpenCode** and **Pi** want their own folders and frontmatter. So your agents
get left behind every time you switch — and you rebuild them by hand. Again.
## ✅ The fix
**Agentic Swarm is a portable agent library.** Define an agent once; the CLI translates it into each
tool's native format and drops it in the right place. Switch tools, set up a new laptop, or onboard a
teammate — your whole swarm comes with you.
| | Without Agentic Swarm | With Agentic Swarm |
|---|---|---|
| New tool | Recreate every agent by hand | `npx … spawn 6` — done |
| Formats | Learn each tool's frontmatter & paths | Handled for you |
| Sharing | "Here, copy-paste these prompts" | One command |
| Source of truth | Scattered across machines | One repo |
---
## 🚀 Quick start
```bash
# 1. See the numbered team list
npx github:hlsitechio/agentic-swarm list
# 2. Spawn a team — by number or name (into Claude Code, the default)
npx github:hlsitechio/agentic-swarm spawn 6 # → security team
npx github:hlsitechio/agentic-swarm spawn backend # → by name
# 3. Spawn into any tool, or just one agent
npx github:hlsitechio/agentic-swarm spawn 3 --target=vscode --project
npx github:hlsitechio/agentic-swarm spawn code-reviewer
# Shortcut: the number/name IS the command
npx github:hlsitechio/agentic-swarm 6
```
Then invoke it: in Claude Code say *“use the code-reviewer agent”*; in VS Code pick it from the Chat
agents dropdown; in OpenCode/Cursor type `@code-reviewer`. Same agent, every tool.
> **No Node installed?** One-liner for Claude Code:
> ```bash
> # macOS / Linux
> curl -fsSL https://raw.githubusercontent.com/hlsitechio/agentic-swarm/main/install.sh | sh -s security
> ```
> ```powershell
> # Windows PowerShell
> irm https://raw.githubusercontent.com/hlsitechio/agentic-swarm/main/install.ps1 | iex; Install-Swarm security
> ```
---
## ⚡ Spawn any team
Copy-paste a command — each spawns the whole team into **Claude Code**. Add `--target=vscode`
(or `codex`, `opencode`, `cursor`, `pi`) for another tool, and `--project` to install into the current repo.
| # | Team | Command |
|---|------|---------|
| 1 | 🏛️ Architecture Guild | `npx github:hlsitechio/agentic-swarm spawn architecture` |
| 2 | ⚙️ Backend Squad | `npx github:hlsitechio/agentic-swarm spawn backend` |
| 3 | 🎨 Frontend Squad | `npx github:hlsitechio/agentic-swarm spawn frontend` |
| 4 | 🔤 Language Pros | `npx github:hlsitechio/agentic-swarm spawn languages` |
| 5 | ✅ Quality Crew | `npx github:hlsitechio/agentic-swarm spawn quality` |
| 6 | 🔐 Security Team | `npx github:hlsitechio/agentic-swarm spawn security` |
| 7 | 🚀 DevOps & SRE | `npx github:hlsitechio/agentic-swarm spawn devops` |
| 8 | 🧠 Data & AI | `npx github:hlsitechio/agentic-swarm spawn data-ai` |
| 9 | 📋 Product & Docs | `npx github:hlsitechio/agentic-swarm spawn product-docs` |
| 10 | 🧩 Specialists | `npx github:hlsitechio/agentic-swarm spawn specialists` |
> **Prefer numbers?** `npx … spawn 1` … `spawn 10` work identically. Spawn several at once: `spawn 2 6`.
> **Spawn everything** (all 70 agents, leads included): `npx … spawn architecture backend frontend languages quality security devops data-ai product-docs specialists`
---
## 🧩 Supported tools
Pick where agents land with `--target` (default: `claude`). Use `--project` for the current repo or
`--global` for your user config.
| Tool | `--target` | Installs to | Invoke with |
|------|-----------|-------------|-------------|
| **Claude Code** | `claude` *(default)* | `~/.claude/agents/` · `.claude/agents/` | auto-delegate · `/agents` |
| **VS Code (Copilot)** | `vscode` | `.github/agents/` · `~/.copilot/agents/` | Chat **agents dropdown** |
| **Codex CLI** | `codex` | `~/.codex/prompts/` | `/prompts:` |
| **OpenCode** | `opencode` | `~/.config/opencode/agents/` · `.opencode/agents/` | `@` |
| **Cursor** | `cursor` | `.cursor/rules/` *(project)* | `@` in chat |
| **Pi** | `pi` | `~/.pi/agent/prompts/` · `.pi/prompts/` | `/` |
| **Anything else** | `generic` | `./agentic-swarm-agents/` | reference the `.md` |
```bash
# Deploy the same team into several tools at once:
npx github:hlsitechio/agentic-swarm spawn security --target=claude,vscode,codex,opencode
```
---
## 🛡️ Guardrails & cost
Every agent ships with a **scoped tool allowlist** and a **model tier** — which keeps the swarm safe *and* cheap:
- **Read-only where it matters.** Auditors, reviewers, architects and planners get only `Read/Grep/Glob` —
they **cannot** edit files or run commands. Engineers get `Edit/Write`; DevOps gets `Bash`. Enforced via
Claude Code `tools:` and OpenCode `permission:`.
- **Big cost cut.** A scoped `tools:` list stops each subagent from loading your **entire MCP catalog** into
its context — often a **3–4× token reduction** per agent (the hidden tax on unrestricted swarms).
- **Model tiers.** Mechanical agents run on **Haiku**, most engineers on **Sonnet**, deep-reasoning agents and
team leads on **Opus** — quality where it counts, cheap where it doesn't.
Retune in [`scripts/generate.py`](scripts/generate.py) (`ACCESS_*`, `TIER_*`) and run `python scripts/generate.py`.
---
## 👥 The roster — 70 agents in 10 teams
Each team has a **`-lead`** that orients on your codebase, dispatches its specialists (parallel or pipeline), and writes one consolidated report — plus **6 specialists**. In Claude Code the lead installs as a **`/-lead` slash command** (it runs in your main session, so it can actually dispatch the specialist subagents — a subagent can't); in OpenCode it's a **primary agent**. Deploy a whole **team** to cover a domain end-to-end, or a single **agent** for a precise task.
Run `list ` to see each agent's description.
**🏛️ `architecture`** — *systems, APIs & boundaries*
`solution-architect` · `api-designer` · `domain-modeler` · `cloud-architect` · `tech-lead` · `integration-architect`
**⚙️ `backend`** — *server-side & data*
`backend-engineer` · `microservices-engineer` · `database-engineer` · `graphql-engineer` · `realtime-engineer` · `caching-engineer`
**🎨 `frontend`** — *UI, client & UX delivery*
`frontend-engineer` · `react-specialist` · `design-system-engineer` · `accessibility-engineer` · `mobile-engineer` · `web-perf-engineer`
**🔤 `languages`** — *deep per-language pros*
`python-pro` · `typescript-pro` · `go-pro` · `rust-pro` · `java-pro` · `csharp-pro`
**✅ `quality`** — *correctness, tests & perf*
`code-reviewer` · `qa-engineer` · `test-automation-engineer` · `performance-engineer` · `refactoring-specialist` · `debugger`
**🔐 `security`** — *appsec, offense & supply chain*
`security-auditor` · `appsec-engineer` · `penetration-tester` · `secrets-scanner` · `dependency-auditor` · `threat-modeler`
**🚀 `devops`** — *ship, scale & operate*
`devops-engineer` · `site-reliability-engineer` · `kubernetes-engineer` · `ci-cd-engineer` · `terraform-engineer` · `observability-engineer`
**🧠 `data-ai`** — *pipelines, models & LLM apps*
`data-engineer` · `ml-engineer` · `data-scientist` · `mlops-engineer` · `analytics-engineer` · `prompt-engineer`
**📋 `product-docs`** — *plan, document & communicate*
`product-manager` · `technical-writer` · `ux-researcher` · `api-documenter` · `release-manager` · `project-planner`
**🧩 `specialists`** — *targeted, high-leverage*
`migration-specialist` · `legacy-modernizer` · `payments-engineer` · `search-engineer` · `i18n-engineer` · `seo-engineer`
---
## 🧰 CLI
```text
npx github:hlsitechio/agentic-swarm [names...] [flags]
Commands
list [team|#] List all teams (numbered), or one team's agents
spawn <#|team|agent> Deploy team(s)/agent(s) — by number, name, or both
remove <#|team|agent> Remove team(s)/agent(s)
<#|team|agent> Shorthand: the number/name IS the command (= spawn)
help Show help
(add = spawn, despawn = remove)
Flags
--target= claude (default), vscode, codex, opencode, cursor, pi, generic
--project Install into ./ (this repo) instead of global config
--global Force global install
--out= Custom output dir for the 'generic' target
--force Overwrite existing files
--dry-run Preview without writing
```
**Examples**
```bash
npx github:hlsitechio/agentic-swarm list security
npx github:hlsitechio/agentic-swarm spawn quality --target=vscode --project
npx github:hlsitechio/agentic-swarm spawn python-pro typescript-pro --target=opencode
npx github:hlsitechio/agentic-swarm remove data-ai
```
---
## ⚙️ How it works
Each agent is stored **once**, as a canonical Markdown file with frontmatter:
```
agents//agent.md # name + description + system prompt
teams/.json # which agents belong to a team
```
When you `add`, an **adapter** per tool rewrites that canonical file into the target's required
format (frontmatter keys, file extension) and writes it to the correct directory — paths verified
against each tool's current docs. **One definition, six outputs.**
Everything is generated from a single source of truth — [`scripts/generate.py`](scripts/generate.py)
defines the full roster and teams, then emits `agents/` and `teams/`.
---
## ❓ FAQ
**Do these replace my tool's built-in agent?** No — they're added alongside, invoked by name.
**Will `add` clobber my existing files?** No. It skips files that already exist unless you pass `--force`.
**Global or project?** Most tools support both. `--project` keeps agents with the repo (commit them,
share with the team); the default is your global user config. Cursor is project-only; Codex is global-only.
**Can I customize the agents?** Yes — fork, edit `scripts/generate.py`, run it, and install from your fork.
---
## 🤝 Contributing
1. Fork this repo
2. Add an entry to `AGENTS` in [`scripts/generate.py`](scripts/generate.py) — `title`, `role`, `does`, `principles`, `use`
3. Add the slug to one or more teams in `TEAMS`
4. Run `python scripts/generate.py` and commit the generated files
5. Open a PR
---
## 🔗 Related
- [**claude-crew**](https://github.com/hlsitechio/claude-crew) — production agent presets for Claude Code
- [**claude-memory**](https://github.com/hlsitechio/claude-memory) — persistent memory for Claude Code sessions
---
**Write once. Deploy anywhere.** · 70 agents · 10 teams + leads · 6 tools · one command.
MIT License