Skip to content
Get packs

Guide · Open source

How to use agency-agents: install and run AI specialist agents

Updated · written with AI from the project's own documentation and checked against our catalogue data

agency-agents is an open-source collection of markdown files, each describing an AI agent with a defined role, personality, processes and deliverables, such as a frontend developer or a Reddit community specialist. This guide explains what the files are, how to install them into tools like Claude Code or Cursor, and how to use them as reference prompts. By the end you will know which install route fits you and how to load one agent for a task.

Agency Agentsmsitarzewski/agency-agents · ★ 166k · MIT

Quick start

  1. 1Pick your route. Choose the desktop app if you want no terminal, or the scripts if you want the command line. Both install the same agents according to the README.
  2. 2Clone the repository. Run git clone on the msitarzewski/agency-agents repository to get the agent files and scripts.
  3. 3Run the installer. Run ./scripts/install.sh with --tool set to your tool, for example claude-code.
  4. 4Limit to a division. Add --division to install only the teams you need, and use --dry-run to preview first.
  5. 5Activate an agent. In your AI tool, name the agent and describe the task in the same message.
  6. 6Review the output. Check the agent's work yourself before relying on it, especially in specialist fields.

What agency-agents is

The Agency is a collection of markdown files. Each file describes one AI agent: its identity and personality traits, core mission and workflows, technical deliverables, success metrics and communication style.

You load the agent you need into a tool such as Claude Code, Cursor or Gemini CLI, and the tool then works in that role. The agents are prompt personalities, so the output depends on the underlying AI tool and on your review of what it produces.

  • Each agent is a markdown file with identity, mission, deliverables and metrics
  • Agents are grouped into divisions, for example engineering and security
  • The project is licensed under MIT

Choose an install route

The README lists several options. The native desktop app, Agency Agents, browses the roster and installs agents into supported tools without cloning the repository. The README says it auto-updates. If you prefer the command line, the shell scripts install the same agents.

The repository is cloned and built from source. The README does not list prerequisites.

  • Desktop app: a native app for macOS, Linux and Windows, released separately from this repository
  • Scripts: clone the repository and run the shell scripts in the scripts folder
  • Reference only: copy and adapt individual agent files

Install all agents into Claude Code

Clone the repository, then run the install script with the claude-code tool option. The script copies the agents into your Claude Code directory.

  • Run the clone command from the README, then move into the project folder
  • Run the install command for Claude Code
git clone https://github.com/msitarzewski/agency-agents ./scripts/install.sh --tool claude-code

Install only one division

If you do not want every agent, install a single division, such as engineering or security. The README also shows a copy method that puts only the engineering category into your Claude Code agents folder.

  • Use --division with a comma-separated list to pick teams
  • Use --dry-run to preview a selection before installing
  • Use --list teams to see every team and its agent count
./scripts/install.sh --tool claude-code --division engineering,security ./scripts/install.sh --tool claude-code --division engineering --dry-run

Use an agent in a session

Once installed, you activate an agent by name in your AI tool. The README gives this example prompt for Claude Code. Describe the task in the same message so the agent has a concrete goal.

  • Pick the agent whose specialty matches the task, using the roster table
  • Review the output yourself, because the agents do not guarantee accuracy
  • For other tools, use the install option for that tool, such as cursor or gemini-cli
Hey Claude, activate Frontend Developer mode and help me build a React component

Use the files as reference prompts

If you are building your own agents, you can read these files and copy the parts you need. Each file follows the same structure, so you can compare them and adapt the sections to your own roles.

  • Read the identity, workflow and deliverables sections of a file
  • Copy and modify the parts that fit your use case

What people use it for

  • Asking a frontend developer agent to help build a React component in Claude Code
  • Installing only the security division for review-focused prompts
  • Copying an agent file as a starting point for a custom role for your team
  • Installing a small set of specialist roles with --agent for a small team

Frequently asked questions

Is agency-agents free to use?

The project is licensed under MIT, according to the README badge and repository licence. The README does not list any paid tier for the agent files.

What are the requirements for agency-agents?

The README does not list prerequisites for the clone-and-build route. You need git to clone the repository and a shell to run the scripts, and the install script targets the tool you name.

How do I install agency-agents into Cursor?

Clone the repository and run ./scripts/install.sh --tool cursor. The README lists cursor among the supported tool options.

Can I install only one team, like engineering?

Yes. Use --division engineering with the install script, or --agent with specific slugs. The README also shows copying the engineering category files directly.

Is there an alternative to the scripts?

The README describes a native desktop app called Agency Agents that installs agents into supported tools with a click. The README does not compare it with other projects.

Are the agents accurate for specialist fields?

The README does not show validation for specialist outputs such as smart contracts, firmware or network configuration. Treat those outputs as drafts and verify them yourself.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

agency-agents suits developers who use Claude Code or Cursor and want role-specific prompts, or teams who want a starting set of specialist roles to adapt. It is not a tested or verified tool, so people who need guaranteed accuracy in specialist fields or a single program with fixed behaviour should choose something else.

Install options & downloads