{"id":50965133,"url":"https://github.com/hlsitechio/agentic-swarm","last_synced_at":"2026-06-18T19:02:15.828Z","repository":{"id":363596875,"uuid":"1161372792","full_name":"hlsitechio/agentic-swarm","owner":"hlsitechio","description":"🐝 Write your AI agents once, deploy anywhere. 60 specialist engineering agents in 10 teams — one command into Claude Code, VS Code, Codex, OpenCode, Cursor \u0026 Pi.","archived":false,"fork":false,"pushed_at":"2026-06-09T16:00:42.000Z","size":216,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-09T16:12:12.637Z","etag":null,"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"],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/hlsitechio.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-02-19T03:08:23.000Z","updated_at":"2026-06-09T15:21:05.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/hlsitechio/agentic-swarm","commit_stats":null,"previous_names":["hlsitechio/agentic-swarm"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/hlsitechio/agentic-swarm","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hlsitechio%2Fagentic-swarm","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hlsitechio%2Fagentic-swarm/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hlsitechio%2Fagentic-swarm/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hlsitechio%2Fagentic-swarm/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/hlsitechio","download_url":"https://codeload.github.com/hlsitechio/agentic-swarm/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hlsitechio%2Fagentic-swarm/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34503511,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-18T02:00:06.871Z","response_time":128,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["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"],"created_at":"2026-06-18T19:02:15.002Z","updated_at":"2026-06-18T19:02:15.816Z","avatar_url":"https://github.com/hlsitechio.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"\u003cdiv align=\"center\"\u003e\n\n\u003cimg src=\"assets/banner.png\" alt=\"Agentic Swarm — AI Agents. United Intelligence.\" width=\"100%\"\u003e\n\n# 🐝 Agentic Swarm\n\n### Write your AI coding agents **once**. Deploy them into **any** tool in seconds.\n\nClaude Code · VS Code · Codex · OpenCode · Cursor · Pi — same agents, one command.\n\n[![Claude Code](https://img.shields.io/badge/Claude_Code-✓-blueviolet?style=flat-square\u0026logo=anthropic)](https://code.claude.com)\n[![VS Code](https://img.shields.io/badge/VS_Code-✓-007ACC?style=flat-square\u0026logo=visualstudiocode)](https://code.visualstudio.com)\n[![Codex](https://img.shields.io/badge/Codex-✓-black?style=flat-square\u0026logo=openai)](https://developers.openai.com/codex)\n[![OpenCode](https://img.shields.io/badge/OpenCode-✓-00bcd4?style=flat-square)](https://opencode.ai)\n[![Cursor](https://img.shields.io/badge/Cursor-✓-blue?style=flat-square)](https://cursor.com)\n[![Pi](https://img.shields.io/badge/Pi-✓-ff5c8a?style=flat-square)](https://pi.dev)\n\u003cbr\u003e\n[![Agents](https://img.shields.io/badge/Agents-70-orange?style=flat-square)]()\n[![Teams](https://img.shields.io/badge/Teams-10-success?style=flat-square)]()\n[![Install](https://img.shields.io/badge/install-npx-cb3837?style=flat-square\u0026logo=npm)]()\n[![License](https://img.shields.io/badge/License-MIT-yellow?style=flat-square)]()\n\n\u003c/div\u003e\n\n---\n\n## 🎯 The problem\n\nYou tune a great set of agents — a sharp code reviewer, a security auditor, a database expert.\nThen you try a new tool, and… you start over.\n\nEvery assistant stores agents differently: **Claude Code** wants `~/.claude/agents/*.md`, **VS Code**\nwants `.github/agents/*.agent.md`, **Cursor** wants `.cursor/rules/*.mdc`, **Codex** wants\n`~/.codex/prompts/`, **OpenCode** and **Pi** want their own folders and frontmatter. So your agents\nget left behind every time you switch — and you rebuild them by hand. Again.\n\n## ✅ The fix\n\n**Agentic Swarm is a portable agent library.** Define an agent once; the CLI translates it into each\ntool's native format and drops it in the right place. Switch tools, set up a new laptop, or onboard a\nteammate — your whole swarm comes with you.\n\n| | Without Agentic Swarm | With Agentic Swarm |\n|---|---|---|\n| New tool | Recreate every agent by hand | `npx … spawn 6` — done |\n| Formats | Learn each tool's frontmatter \u0026 paths | Handled for you |\n| Sharing | \"Here, copy-paste these prompts\" | One command |\n| Source of truth | Scattered across machines | One repo |\n\n---\n\n## 🚀 Quick start\n\n```bash\n# 1. See the numbered team list\nnpx github:hlsitechio/agentic-swarm list\n\n# 2. Spawn a team — by number or name (into Claude Code, the default)\nnpx github:hlsitechio/agentic-swarm spawn 6          # → security team\nnpx github:hlsitechio/agentic-swarm spawn backend    # → by name\n\n# 3. Spawn into any tool, or just one agent\nnpx github:hlsitechio/agentic-swarm spawn 3 --target=vscode --project\nnpx github:hlsitechio/agentic-swarm spawn code-reviewer\n\n# Shortcut: the number/name IS the command\nnpx github:hlsitechio/agentic-swarm 6\n```\n\nThen invoke it: in Claude Code say *“use the code-reviewer agent”*; in VS Code pick it from the Chat\nagents dropdown; in OpenCode/Cursor type `@code-reviewer`. Same agent, every tool.\n\n\u003e **No Node installed?** One-liner for Claude Code:\n\u003e ```bash\n\u003e # macOS / Linux\n\u003e curl -fsSL https://raw.githubusercontent.com/hlsitechio/agentic-swarm/main/install.sh | sh -s security\n\u003e ```\n\u003e ```powershell\n\u003e # Windows PowerShell\n\u003e irm https://raw.githubusercontent.com/hlsitechio/agentic-swarm/main/install.ps1 | iex; Install-Swarm security\n\u003e ```\n\n---\n\n## ⚡ Spawn any team\n\nCopy-paste a command — each spawns the whole team into **Claude Code**. Add `--target=vscode`\n(or `codex`, `opencode`, `cursor`, `pi`) for another tool, and `--project` to install into the current repo.\n\n| # | Team | Command |\n|---|------|---------|\n| 1 | 🏛️ Architecture Guild | `npx github:hlsitechio/agentic-swarm spawn architecture` |\n| 2 | ⚙️ Backend Squad | `npx github:hlsitechio/agentic-swarm spawn backend` |\n| 3 | 🎨 Frontend Squad | `npx github:hlsitechio/agentic-swarm spawn frontend` |\n| 4 | 🔤 Language Pros | `npx github:hlsitechio/agentic-swarm spawn languages` |\n| 5 | ✅ Quality Crew | `npx github:hlsitechio/agentic-swarm spawn quality` |\n| 6 | 🔐 Security Team | `npx github:hlsitechio/agentic-swarm spawn security` |\n| 7 | 🚀 DevOps \u0026 SRE | `npx github:hlsitechio/agentic-swarm spawn devops` |\n| 8 | 🧠 Data \u0026 AI | `npx github:hlsitechio/agentic-swarm spawn data-ai` |\n| 9 | 📋 Product \u0026 Docs | `npx github:hlsitechio/agentic-swarm spawn product-docs` |\n| 10 | 🧩 Specialists | `npx github:hlsitechio/agentic-swarm spawn specialists` |\n\n\u003e **Prefer numbers?** `npx … spawn 1` … `spawn 10` work identically. Spawn several at once: `spawn 2 6`.\n\u003e **Spawn everything** (all 70 agents, leads included): `npx … spawn architecture backend frontend languages quality security devops data-ai product-docs specialists`\n\n---\n\n## 🧩 Supported tools\n\nPick where agents land with `--target` (default: `claude`). Use `--project` for the current repo or\n`--global` for your user config.\n\n| Tool | `--target` | Installs to | Invoke with |\n|------|-----------|-------------|-------------|\n| **Claude Code** | `claude` *(default)* | `~/.claude/agents/` · `.claude/agents/` | auto-delegate · `/agents` |\n| **VS Code (Copilot)** | `vscode` | `.github/agents/` · `~/.copilot/agents/` | Chat **agents dropdown** |\n| **Codex CLI** | `codex` | `~/.codex/prompts/` | `/prompts:\u003cagent\u003e` |\n| **OpenCode** | `opencode` | `~/.config/opencode/agents/` · `.opencode/agents/` | `@\u003cagent\u003e` |\n| **Cursor** | `cursor` | `.cursor/rules/` *(project)* | `@\u003cagent\u003e` in chat |\n| **Pi** | `pi` | `~/.pi/agent/prompts/` · `.pi/prompts/` | `/\u003cagent\u003e` |\n| **Anything else** | `generic` | `./agentic-swarm-agents/` | reference the `.md` |\n\n```bash\n# Deploy the same team into several tools at once:\nnpx github:hlsitechio/agentic-swarm spawn security --target=claude,vscode,codex,opencode\n```\n\n---\n\n## 🛡️ Guardrails \u0026 cost\n\nEvery agent ships with a **scoped tool allowlist** and a **model tier** — which keeps the swarm safe *and* cheap:\n\n- **Read-only where it matters.** Auditors, reviewers, architects and planners get only `Read/Grep/Glob` —\n  they **cannot** edit files or run commands. Engineers get `Edit/Write`; DevOps gets `Bash`. Enforced via\n  Claude Code `tools:` and OpenCode `permission:`.\n- **Big cost cut.** A scoped `tools:` list stops each subagent from loading your **entire MCP catalog** into\n  its context — often a **3–4× token reduction** per agent (the hidden tax on unrestricted swarms).\n- **Model tiers.** Mechanical agents run on **Haiku**, most engineers on **Sonnet**, deep-reasoning agents and\n  team leads on **Opus** — quality where it counts, cheap where it doesn't.\n\nRetune in [`scripts/generate.py`](scripts/generate.py) (`ACCESS_*`, `TIER_*`) and run `python scripts/generate.py`.\n\n---\n\n## 👥 The roster — 70 agents in 10 teams\n\nEach team has a **`\u003cteam\u003e-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 **`/\u003cteam\u003e-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.\nRun `list \u003cteam|#\u003e` to see each agent's description.\n\n\u003ctable\u003e\n\u003ctr\u003e\u003ctd valign=\"top\" width=\"50%\"\u003e\n\n**🏛️ `architecture`** — *systems, APIs \u0026 boundaries*\n`solution-architect` · `api-designer` · `domain-modeler` · `cloud-architect` · `tech-lead` · `integration-architect`\n\n**⚙️ `backend`** — *server-side \u0026 data*\n`backend-engineer` · `microservices-engineer` · `database-engineer` · `graphql-engineer` · `realtime-engineer` · `caching-engineer`\n\n**🎨 `frontend`** — *UI, client \u0026 UX delivery*\n`frontend-engineer` · `react-specialist` · `design-system-engineer` · `accessibility-engineer` · `mobile-engineer` · `web-perf-engineer`\n\n**🔤 `languages`** — *deep per-language pros*\n`python-pro` · `typescript-pro` · `go-pro` · `rust-pro` · `java-pro` · `csharp-pro`\n\n**✅ `quality`** — *correctness, tests \u0026 perf*\n`code-reviewer` · `qa-engineer` · `test-automation-engineer` · `performance-engineer` · `refactoring-specialist` · `debugger`\n\n\u003c/td\u003e\u003ctd valign=\"top\" width=\"50%\"\u003e\n\n**🔐 `security`** — *appsec, offense \u0026 supply chain*\n`security-auditor` · `appsec-engineer` · `penetration-tester` · `secrets-scanner` · `dependency-auditor` · `threat-modeler`\n\n**🚀 `devops`** — *ship, scale \u0026 operate*\n`devops-engineer` · `site-reliability-engineer` · `kubernetes-engineer` · `ci-cd-engineer` · `terraform-engineer` · `observability-engineer`\n\n**🧠 `data-ai`** — *pipelines, models \u0026 LLM apps*\n`data-engineer` · `ml-engineer` · `data-scientist` · `mlops-engineer` · `analytics-engineer` · `prompt-engineer`\n\n**📋 `product-docs`** — *plan, document \u0026 communicate*\n`product-manager` · `technical-writer` · `ux-researcher` · `api-documenter` · `release-manager` · `project-planner`\n\n**🧩 `specialists`** — *targeted, high-leverage*\n`migration-specialist` · `legacy-modernizer` · `payments-engineer` · `search-engineer` · `i18n-engineer` · `seo-engineer`\n\n\u003c/td\u003e\u003c/tr\u003e\n\u003c/table\u003e\n\n---\n\n## 🧰 CLI\n\n```text\nnpx github:hlsitechio/agentic-swarm \u003ccommand\u003e [names...] [flags]\n\nCommands\n  list [team|#]        List all teams (numbered), or one team's agents\n  spawn \u003c#|team|agent\u003e Deploy team(s)/agent(s) — by number, name, or both\n  remove \u003c#|team|agent\u003e Remove team(s)/agent(s)\n  \u003c#|team|agent\u003e       Shorthand: the number/name IS the command (= spawn)\n  help                 Show help\n                       (add = spawn, despawn = remove)\n\nFlags\n  --target=\u003ct[,t...]\u003e  claude (default), vscode, codex, opencode, cursor, pi, generic\n  --project            Install into ./ (this repo) instead of global config\n  --global             Force global install\n  --out=\u003cdir\u003e          Custom output dir for the 'generic' target\n  --force              Overwrite existing files\n  --dry-run            Preview without writing\n```\n\n**Examples**\n\n```bash\nnpx github:hlsitechio/agentic-swarm list security\nnpx github:hlsitechio/agentic-swarm spawn quality --target=vscode --project\nnpx github:hlsitechio/agentic-swarm spawn python-pro typescript-pro --target=opencode\nnpx github:hlsitechio/agentic-swarm remove data-ai\n```\n\n---\n\n## ⚙️ How it works\n\nEach agent is stored **once**, as a canonical Markdown file with frontmatter:\n\n```\nagents/\u003cslug\u003e/agent.md     # name + description + system prompt\nteams/\u003cid\u003e.json            # which agents belong to a team\n```\n\nWhen you `add`, an **adapter** per tool rewrites that canonical file into the target's required\nformat (frontmatter keys, file extension) and writes it to the correct directory — paths verified\nagainst each tool's current docs. **One definition, six outputs.**\n\nEverything is generated from a single source of truth — [`scripts/generate.py`](scripts/generate.py)\ndefines the full roster and teams, then emits `agents/` and `teams/`.\n\n---\n\n## ❓ FAQ\n\n**Do these replace my tool's built-in agent?** No — they're added alongside, invoked by name.\n\n**Will `add` clobber my existing files?** No. It skips files that already exist unless you pass `--force`.\n\n**Global or project?** Most tools support both. `--project` keeps agents with the repo (commit them,\nshare with the team); the default is your global user config. Cursor is project-only; Codex is global-only.\n\n**Can I customize the agents?** Yes — fork, edit `scripts/generate.py`, run it, and install from your fork.\n\n---\n\n## 🤝 Contributing\n\n1. Fork this repo\n2. Add an entry to `AGENTS` in [`scripts/generate.py`](scripts/generate.py) — `title`, `role`, `does`, `principles`, `use`\n3. Add the slug to one or more teams in `TEAMS`\n4. Run `python scripts/generate.py` and commit the generated files\n5. Open a PR\n\n---\n\n## 🔗 Related\n\n- [**claude-crew**](https://github.com/hlsitechio/claude-crew) — production agent presets for Claude Code\n- [**claude-memory**](https://github.com/hlsitechio/claude-memory) — persistent memory for Claude Code sessions\n\n\u003cdiv align=\"center\"\u003e\n\n---\n\n**Write once. Deploy anywhere.** · 70 agents · 10 teams + leads · 6 tools · one command.\n\nMIT License\n\n\u003c/div\u003e\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhlsitechio%2Fagentic-swarm","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fhlsitechio%2Fagentic-swarm","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhlsitechio%2Fagentic-swarm/lists"}