# AI Agent Types

Agentic Swarm describes thirteen specialized agent roles grouped across the six layers of Organizational Intelligence.

## 01. Governor Agent

The Governor Agent enforces system policies, risk limits, and safety rules. It can approve, block, or halt any agent action before it executes, and it decides when a human must sign off.

**Why it matters:** It is the single point that keeps autonomy inside approved boundaries. Without it, a capable swarm can act faster than anyone can review.

[Read the Governor Agent page](/agents/governor-agent.md)

## 02. Monitor Agent

The Monitor Agent watches the whole system for performance, cost, safety, loops, and failures. It surfaces problems early and can trigger a halt when something goes wrong.

**Why it matters:** It gives governance and learning the live signals they need. Problems caught early stay small.

[Read the Monitor Agent page](/agents/monitor-agent.md)

## 03. Planner Agent

The Planner Agent converts goals into executable plans. It selects strategies, defines workflows, and adapts the plan when conditions change.

**Why it matters:** A mission without a plan cannot be coordinated. The planner gives every other agent a shared structure to work from.

[Read the Planner Agent page](/agents/planner-agent.md)

## 04. Routing Agent

The Routing Agent selects the best agents, models, tools, and workflows for each task, optimizing for cost, speed, and quality.

**Why it matters:** Where a task is sent decides how well and how cheaply it gets done. Routing keeps the system efficient as it scales.

[Read the Routing Agent page](/agents/routing-agent.md)

## 05. Manager Agent

The Manager Agent orchestrates execution across specialized agents. It manages dependencies, tracks progress, and ensures objectives are completed.

**Why it matters:** Plans do not run themselves. The manager keeps parallel work aligned and moving toward the mission.

[Read the Manager Agent page](/agents/manager-agent.md)

## 06. Memory Agent

The Memory Agent manages short-term, long-term, and compressed memory. It controls what the system stores and what it retrieves when needed.

**Why it matters:** Memory is what lets a loop start smarter than the last. It turns scattered runs into institutional knowledge.

[Read the Memory Agent page](/agents/memory-agent.md)

## 07. Evaluator Agent

The Evaluator Agent critiques outputs and improves them through feedback loops. It checks work against the brief and either refines it or sends it back.

**Why it matters:** It keeps quality high as volume grows, and the lessons it produces are what the learning layer stores. It sits in the Memory and Recursive Learning Layer because evaluation is how the system learns what good looks like.

[Read the Evaluator Agent page](/agents/evaluator-agent.md)

## 08. Specialist Agent

Specialist Agents are domain-specific workers operating under a manager. Each focuses on a single area of expertise, such as support, coding, research, sales, or legal review.

**Why it matters:** Depth beats breadth on real work. Specialists produce higher-quality output in their domain than a single general agent could.

[Read the Specialist Agent page](/agents/specialist-agent.md)

## 09. Tool-Using Agent

The Tool-Using Agent executes explicit functions through tools and APIs. It reads, writes, and acts on external systems on behalf of the swarm.

**Why it matters:** It is how intent becomes action in the real world. Without tool use, agents can only talk, not do.

[Read the Tool-Using Agent page](/agents/tool-using-agent.md)

## 10. Function-Triggering Agent

The Function-Triggering Agent emits structured outputs that trigger predefined external functions. It produces the exact payload a downstream system expects.

**Why it matters:** It connects agent reasoning to reliable, predefined actions, so high-stakes steps run the same way every time.

[Read the Function-Triggering Agent page](/agents/function-triggering-agent.md)

## 11. Connector Agent

The Connector Agent uses sandboxed extensions to access external services safely. It reaches outside systems without exposing the core swarm to them.

**Why it matters:** It extends what the swarm can reach while keeping external access contained and governed.

[Read the Connector Agent page](/agents/connector-agent.md)

## 12. Delivery Agent

The Delivery Agent packages outputs into polished deliverables for humans or clients. It formats finished work for its audience and hands it off.

**Why it matters:** Value only lands when work is delivered in a usable form. The delivery agent is the last mile of every mission.

[Read the Delivery Agent page](/agents/delivery-agent.md)

## 13. Simulation Agent

The Simulation Agent models possible outcomes and tests strategies before execution. It runs scenarios so the swarm can choose the strongest path first.

**Why it matters:** It lets the system think before it acts. Testing options in simulation is cheaper and safer than learning only in production.

[Read the Simulation Agent page](/agents/simulation-agent.md)

---

Canonical HTML: https://agenticswarm.com/agents