Introduction to Agents

MeshAgent agents are room-connected AI participants that can talk to people, call tools, work with shared room state, and stay available as services.

In MeshAgent, an agent is more than a model prompt. A MeshAgent agent can join a room, receive work, use tools, read and write shared data, and surface its work through chat, voice, mail, queues, or tool calls depending on how you configure it.

What agents can do

MeshAgent agents can be used to:

• answer questions in threaded conversations
• search the web, read files, and write documents
• process background jobs from queues
• handle inbound email
• expose themselves as tools for other agents or applications
• work with room storage, database tables, memory, and custom toolkits
• stay available as room or project services instead of only running as one-off scripts

This is what makes MeshAgent useful for support agents, coding agents, research agents, operations agents, and other workflows that need tools, persistence, and deployment.

How to build agents with MeshAgent

MeshAgent supports two main ways to build agents:

Use the CLI

The CLI is the fastest way to get a production-ready agent running in a room.

For most new text and background agents, start with meshagent process.

Use the SDKs

Use SingleRoomAgent and the MeshAgent SDKs when you need custom room-connected behavior or deeper control over agent logic, tool resolution, or service composition.

What MeshAgent provides around your agent

MeshAgent handles the runtime pieces around the agent logic you write:

• room connectivity and identity
• tool and toolkit discovery
• threaded conversations and persisted thread state
• UI integration through MeshAgent Studio and SDK clients
• deployment as room or project services

That means you can focus on the agent’s rules, tools, and behavior instead of building the surrounding infrastructure yourself.

A simple deployment example

Once a process-backed agent works locally, you can deploy it directly from the CLI:

meshagent process deploy \
  --service-name support-agent \
  --room quickstart \
  --agent-name support-agent \
  --channel chat \
  --channel queue:support-jobs \
  --channel toolkit:support-agent \
  --web-search \
  --require-storage \
  --rule "You are a helpful support agent."

This deploys a room service that runs the same kind of process-backed agent you would test locally with meshagent process join.

Older patterns

Older dedicated runtimes and older SDK patterns are being phased out of the main docs. For new text and background agents, prefer meshagent process.

Running and deploying agents

MeshAgent agents can be run locally during development and deployed later as room or project services.

Mode	Description	Best For
Process Agents Overview	Run a process-backed agent directly from your machine with the MeshAgent CLI.	Testing and iteration
Project Service	Deploy once to run automatically in every room of a project.	Shared always-on agents
Room Service	Deploy to a specific room only.	Room-scoped agents

Next Steps

• Process Agents Overview: Learn what meshagent process is, what it supports, and how to run it.
• Process Agent Architecture: Understand how process-backed agents work behind the scenes.
• VoiceBot: Learn the current dedicated voice runtime.
• SingleRoomAgent: Build a custom room-connected agent with the SDK.