Home / Hire AI Agents
AI Workforce

Hire AI Agents Like Employees

Give each agent a role, a job description, and a task. They read your codebase, write production code, and produce structured handoff documents when done. Fire them, archive them, rehire the best ones.

Hire Your First Agent Setup Guide
The AI Workforce Model

Treat AI agents like your engineering team

chAIrman uses the employment metaphor because it maps perfectly to how AI agents should work. You don't just prompt an AI. You hire it for a specific role, give it a job description that sets expectations, assign tasks with clear deliverables, monitor its performance, and fire it when the job is done.

This structure prevents the chaos of unmanaged AI. Every agent has a defined scope. Every task has success criteria. Every completed task produces a handoff document. When you fire an agent, their knowledge is archived. When you need that role again, you rehire the veteran.

  • Hire: Define role, job description, model (Sonnet or Opus), and max retries
  • Assign: Give tasks with file ownership, dependencies, and success criteria
  • Monitor: Real-time dashboard, desktop notifications, cost tracking
  • Fire: Archive to alumni with full history. Rehire veterans anytime.
// Hire a specialized agent hire_agent({ project: "my-saas", role: "Security Auditor", job_description: ` Audit all API routes for OWASP Top 10. Focus on auth middleware, input validation, and SQL injection vectors. Tech: Node.js, Express, PostgreSQL. Read: src/routes/, src/middleware/ Do NOT modify any files. Output: security-report.md `, model: "opus" }) // Agent is hired, ready for task // Status: "hired"
Pre-Built Roles

10 agent templates ready to hire

Each template comes with a comprehensive job description. Just specify your tech stack and project path.

FE

Frontend Lead

Builds UI components, pages, and client-side logic. Handles responsive design, accessibility, state management, and API integration. Recommended model: Sonnet.

BE

Backend Lead

Builds APIs, server logic, database queries, and integrations. Handles authentication, authorization, error handling, and data validation. Recommended model: Sonnet.

QA

QA Engineer

Writes unit tests, integration tests, and E2E tests. Validates code against success criteria. Catches edge cases and regression bugs. Recommended model: Sonnet.

DB

Database Engineer

Designs schemas, writes migrations, optimizes queries, and manages indexes. Handles data modeling, relationships, and constraints. Recommended model: Sonnet.

SEC

Security Auditor

Audits code for OWASP Top 10 vulnerabilities. Reviews auth flows, input validation, encryption, and access controls. Produces detailed security reports. Recommended model: Opus.

OPS

DevOps Engineer

Configures CI/CD pipelines, Docker containers, deployment scripts, and monitoring. Handles environment configuration and infrastructure. Recommended model: Sonnet.

Plus: API Designer, Code Reviewer, Documentation Writer, and Performance Engineer templates.

Institutional Memory

Fire agents, keep their knowledge

When you fire an agent, chAIrman doesn't just kill the process. It archives the agent to the alumni roster with their complete history: every task completed, every file changed, total cost, success rate, and structured handoff notes.

When you need that role again, run rehire_veteran to bring back a proven agent. The new agent inherits the veteran's job description and experience notes. It already knows your codebase patterns, what worked, and what didn't.

  • Alumni archive stores: role, job description, model, task history, files changed, cost, success rate
  • Rehired veterans skip the learning curve and start producing from day one
  • Your project accumulates institutional knowledge that compounds over time
  • View alumni performance stats to identify your best agents for rehiring
// Check your alumni roster list_alumni({ project: "my-saas" }) // Returns: // ID Role Tasks Rate Cost // alum-1 Backend Lead 8 100% $4.20 // alum-2 Frontend Lead 5 80% $2.15 // alum-3 QA Engineer 6 100% $1.80 // Rehire the best backend veteran rehire_veteran({ project: "my-saas", archive_id: "alum-1" }) // New agent inherits: // - Original job description // - Experience notes from 8 tasks // - Knowledge of your codebase patterns
10
Agent Templates
6
Agent Statuses
857
Skills Injected
3
Auto-Retries
Common Questions

Hiring AI agents FAQ

What does it mean to "hire" an AI agent?
Hiring an AI agent in chAIrman creates an autonomous Claude Code process with a defined role, job description, and model assignment. The agent receives your project's CLAUDE.md, relevant skills from the library, and any team messages. It's ready to receive tasks. Think of it as onboarding a contractor who already knows your tech stack.
How do I write a good job description?
A good job description includes: the tech stack, specific file paths the agent should work with, files that are off-limits, coding conventions to follow, and what "done" looks like. Be as specific as you would with a human contractor. The more context you provide, the better the agent performs. Or use a pre-built template and just specify your tech stack.
Can I hire multiple agents for the same project?
Yes. chAIrman is designed for multi-agent workflows. Hire as many agents as you need. The ticket system ensures no two agents edit the same file simultaneously. Use dependency pipelines to order work correctly. The default concurrency limit is 10 agents, configurable via CHAIRMAN_MAX_AGENTS.
What happens when I fire an agent?
The agent's process is terminated, a handoff document is saved, and the agent is archived to the alumni roster. The archive includes: role, job description, model used, tasks completed, files changed, total cost, and success rate. You can rehire veterans later to get their institutional knowledge back.
How much does each agent cost?
Agent costs depend on the Claude model used and task complexity. Sonnet-class agents typically cost $0.20-$0.80 per task. Opus-class agents cost $1-$5 for complex tasks. chAIrman tracks costs per-agent, per-task, and per-project. Set a project budget to prevent overspending.
Can agents communicate with each other?
Yes. Agents can broadcast messages to the team or send direct messages to specific agents using output tags. The CEO can also post messages via the message board. Messages are injected into agent prompts at task-assignment time. For real-time coordination, use dependency pipelines to ensure work happens in the right order.

Ready to build your AI engineering team?

Hire agents, assign tasks, ship features. Your AI workforce starts at $19.99/mo.

Start Hiring Setup Guide