Agency Agents is a collection of specialized AI agent personas for coding assistants. Each persona is a Markdown file that sets out one domain expert: its identity and communication style, its core mission and workflow, the technical deliverables it produces with code examples, and the success metrics it works to. The roster is organized into divisions such as engineering, design, product, marketing, sales, testing, security, finance, game development, healthcare, and research.
The agents install natively into Claude Code as .md files, and conversion and install scripts rewrite the same personas into each other supported tool’s native layout — SKILL.md skills, .mdc rule files, a single CONVENTIONS.md, TOML agent definitions, or YAML specs — then copy them where that tool expects them. An interactive installer detects which tools you have and lets you take the whole roster, one or more divisions, or named agents; a --dry-run flag previews a selection first.
A separate desktop application for macOS, Linux, and Windows browses the roster and performs the same installation without cloning the repository or running scripts. Because some runtimes cap how many agents they register, the installer warns when a selection would exceed a tool’s documented limit and suggests narrowing it to a division.
Features
- Persona files: each agent is a Markdown definition covering identity, mission, workflow, deliverables, and success metrics
- Division structure: the roster is grouped into engineering, design, product, marketing, sales, testing, security, finance, and other divisions
- Native Claude Code agents: files drop straight into the Claude Code agents directory with no conversion step
- Cross-tool conversion: a conversion script emits the format each supported tool expects, from skill files to rule files to TOML and YAML specs
- Selective installation: install everything, pick divisions, or name individual agents, with a dry-run preview
- Interactive installer: auto-detects the coding tools present on the machine and offers the matching targets
- Registration limits: the installer warns when a selection exceeds a tool’s documented agent-registration cap
- Desktop app: a native macOS, Linux, and Windows app browses the roster and installs agents without the command line
- Worked scenarios: example multi-agent workflows cover cases such as building a startup MVP, launching a marketing campaign, and enterprise feature development
Runs in
Aider
Command-line LLM pair programmer that maps your repo, edits files, and auto-commits each change to git
Claude Code
Terminal-based agentic coding tool that reads your codebase and runs tasks through natural-language commands
Codex
Lightweight coding agent that runs in your terminal
Hermes
Self-improving AI agent with a learning loop that creates and refines skills, recalls past sessions, and runs across chat platforms