Learn
AI agents, explained plainly
Short, direct answers to the questions teams ask before they put AI agents to work: what agents are, how to keep them safe, and how they connect to everything else. Each one says what MyDeepAgents does today, and what it does not do yet.
16 articles · Updated
The basics
What an agent is, when a workflow is the better tool, and how to pick the model.
- Agents5 min read
What is an AI agent, and how is it different from a chatbot?
An AI agent uses a language model to choose and call tools until a goal is met. How agents work, how they differ from chatbots, and when to use one.
Read the answer - Agents5 min read
AI agents vs workflows: when to use which
Use a workflow when you know the steps, and an AI agent when the steps depend on what it finds. A comparison, examples, and how to combine the two.
Read the answer - Agents5 min read
How to choose a model for an AI agent
Test a few models on your own tasks and weigh tool calling, instruction following, cost, speed and where your data may go. Hosted, cloud and local options compared.
Read the answer
Trust and control
Approvals, security controls, and keeping many customers apart.
- Trust and control5 min read
Human-in-the-loop approvals for AI agents
A human-in-the-loop approval makes an AI agent stop before a chosen action until a person says yes. Which actions need one, and how approvals work.
Read the answer - Trust and control6 min read
AI agent security: guardrails, egress rules and audit
The main security risks of AI agents and the controls that answer them: least privilege, approvals, data redaction, egress rules, limits and an audit trail.
Read the answer - Trust and control5 min read
Multi-tenant AI agents: keeping each customer’s data, keys and limits apart
How a multi-tenant AI agent platform keeps each customer’s data, credentials, settings and spending separate: row-level security, per-tenant keys, budgets and delegation.
Read the answer
Open protocols
How agents reach tools over MCP, and other agents over A2A.
- Protocols6 min read
What is MCP (Model Context Protocol)? How AI agents use MCP servers
MCP is an open standard that lets AI apps and agents list and call tools from any MCP server. How it works, how agents use it, and how to serve your own tools.
Read the answer - Protocols5 min read
What is A2A, the Agent2Agent protocol?
A2A (Agent2Agent) is an open protocol that lets AI agents from different companies find each other, send work and get results. How it works, and how it differs from MCP.
Read the answer
Use cases
What teams put agents to work on, and how each one is set up today.
- Use cases5 min read
AI agent use cases: what teams put agents to work on
Eight practical AI agent use cases, from customer support and inbox triage to reports, database questions and team chat, and how to pick your first one.
Read the answer - Use cases4 min read
How to use an AI agent for customer support
Use an AI agent to answer customers in your website chat and to draft replies to help-desk tickets, with orders looked up live and approvals before anything is sent.
Read the answer - Use cases4 min read
How to build an AI email assistant that sorts your inbox
An AI email assistant that reads new mail on a schedule, labels and archives it, posts a summary to Slack and replies only with your approval. For Gmail and Outlook.
Read the answer - Use cases4 min read
How to automate reports with AI: daily and weekly summaries
Have AI pull your numbers on a schedule, explain what changed in plain words and post the summary to Slack, Teams or email. With a workflow, or an agent when it must dig.
Read the answer - Use cases4 min read
How to ask your database questions in plain English with AI
Let an AI agent turn questions into read-only SQL against PostgreSQL, MySQL, SQL Server and more, so your team gets answers without writing queries or risking the data.
Read the answer - Use cases4 min read
A private AI chat for your team, with your models, agents and tools
Give your team one private AI chat that uses your own model keys, searches the web with sources, works with your agents and tools, and asks before it changes anything.
Read the answer
Putting agents to work
Two of the most common first projects, step by step.
- Putting agents to work5 min read
How to add an AI agent to your website
Build the agent, create a chat widget, allow your domain and paste one script tag. Step by step, with visitor sign-in, cost limits and what to watch for.
Read the answer - Putting agents to work4 min read
How to connect an AI agent to your CRM
Connect the CRM once, give the agent only the actions it needs, and make writes wait for approval. Step by step, with HubSpot and Salesforce examples.
Read the answer
Early access
Put an agent to work
Build an agent, connect it to your tools, and decide what it may do on its own. Or tell us what you want it to do.