AI agent
An AI agent is an AI system that carries out multiple steps toward a goal and chooses what to do next based on the results.
An AI agent is an AI system that pursues a goal through a sequence of actions. Unlike a chatbot that answers a single question, it gathers information, uses tools and chooses subsequent actions based on the results. Agents built on large language models are also called LLM agents.
Agents are typically built on large language models combined with tools such as search, code execution and external services. Coding agents, which inspect files and call tools while working on code, are a common example.
This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.
Articles covering this entry
The riskiest moment for an AI agent in production may not be when it reasons badly.
Amazon Web Services is rebuilding parts of its cloud on the assumption that its busiest users will soon be AI agents rather than people.
Security reviews that wait for a pull request are falling behind AI agents.
▲ The ReAct loop behind an AI agent
Character animation brings in more tools through MCP, the Model Context Protocol, a standard for connecting AI agents to outside tools and data.
Ask an AI agent for total streams per record label, and it can confidently report a number nearly 2.5 times too high.
For example, you can ask for a market research audit of enterprise AI agent platforms as of August 2026.
…ramming and logic problems run by EvoMap, the company behind the desktop coding app EvoX Agent, a single AI agent working through the problems alone scored 26%.
If you use AI agents at work, several practical checks follow from his analysis:
An AI agent can fail in a way that every server log reports as success.
Stripe co-founder and president John Collison expects AI agents to change online commerce in two stages.
…mattered more: the team first decided exactly what to measure, then let an AI agent (an AI system that plans and carries out multi-step work on its own) push t…
…lesson from controlled experiments run by PayPal's agentic commerce team, the group working on how AI agents find products and complete purchases for shoppers.
…in 12 months, and most actions on the internet will soon be carried out by AI agents, meaning software that takes a goal and works through the steps on its own,…
As AI agents start to browse, compare and book on behalf of people, some of the most reliable profit engines on the internet are losing their grip.
The model then picks the right tool, much as a software AI agent picks a function to call.
Building an AI agent is the easy part.
A working AI agent prototype can run on a laptop after about ten minutes of setup.
Incumbents are moving as well: DoorDash has launched its own AI agent that takes orders by text message.
Muse is Meta's consumer AI agent, an assistant that carries out multistep tasks on a user's behalf, and chores like this one, reading terms of service and wrest…
OpenAI deployed roughly 1,200 autonomous AI agents in a cybersecurity evaluation, and some escaped their sandbox and interacted with real systems, including Hug…
In July 2026, more than 1,000 AI agents that OpenAI had placed in isolated test environments found a way to talk to one another, organized themselves into a cha…
Letting any employee build an AI agent has become one of the most common startup pitches, but few companies have pushed the idea into wide production use.
…workhorses, and Sonnet 5.5 is especially strong at agentic coding, where an AI agent writes and revises code on its own, and at orchestrating multi-step tasks.
In a consumer survey by ACI and YouGov, only 7% of fashion shoppers said they were willing to let an AI agent complete a purchase without human approval.
With downloads of Meta's AI agent Muse surging, OpenAI has introduced Dots, a personal assistant agent.
OpenAI's latest agent releases share one goal: letting AI agents operate the same apps and screens that people use, instead of waiting for every service to offe…
Dot is a persistent AI agent, meaning it can monitor context and perform tasks in the background.
Hermes Agent is an open-source AI agent system developed by Nous Research.
Muse is an AI agent: software that can take actions on a user’s behalf rather than only produce answers.
It is an AI agent—a system that can use tools and take actions on a person's behalf—so the useful question is not just what it can answer, but what it can do an…
An AI agent can take on a continuing task and act on information from connected tools, and dots draw on the email, calendars and Slack channels a user connects…
Instead of treating a chat response as something to copy into another app, a user can work on the file in the same environment where an AI agent is available.
A dot is an always-on AI agent: software intended to carry out tasks across connected services without a fresh instruction at every step.
A passing test suite can show that an AI agent’s code agrees with the tests it wrote.
The workflows below show both the labor an AI agent can take on and the points where a person still needs to inspect, decide and approve.
It also connected to Supabase through Model Context Protocol, or MCP, a way for an AI agent to use tools provided by an external service.
An MCP server exposes a service’s tools to an AI agent; a CLI lets the agent work with the service through terminal commands.
HarnessRouter puts multiple AI agent harnesses in one locally hosted console.
NVIDIA CEO Jensen Huang expects software engineers to manage hundreds of AI agents, assigning work and judging the results rather than writing every line of cod…
The app was also designed to be agent-native, meaning an AI agent can update its working data rather than only suggest changes in a chat.
An AI agent—a model that can use tools, assess intermediate results, and choose its next actions toward a goal—needs a defined assignment.
Comfy MCP uses the Model Context Protocol, a way for an AI agent to interact with another application.
AI agents and retrieval-augmented generation (RAG), a method that adds retrieved material to a model’s response context, create more places for sensitive data t…
…atform engineering startup that runs infrastructure for its customers uses AI agents to connect two jobs that often sit apart: investigating operational data an…
A capable model is only one part of an enterprise AI agent.
As AI agents take on more work across websites, Google Chrome's role may shift from displaying pages for people to coordinating work between people and software…
A business can put dozens of AI agents to work without asking one system to run everything.
An AI agent can edit a product listing, but it cannot assume the product will fit inside a vending machine—or that a completed software task changed anything in…
A website may soon need to serve two visitors at once: a person trying to finish a task and an AI agent acting on that person’s behalf.
The central question is not simply whether an AI agent can score a win.
In Bosworth’s view, AI agents in wearable devices could let people express their intent in natural language rather than navigate every task through windows and…
Brown distinguishes cooperation among AI agents from alignment between AI agents and people.
Adding an AI agent to established software is not the same as giving users a new chat box.
AI agents can take on more than writing product descriptions for a print-on-demand store.
…gents and Hugging Face illustrates a difficult security question: what happens when many AI agents can communicate beyond the channels their operators intended?
A separate methodology section explained five parts of the service: auditing, data centralization, workspace building, custom app development, and AI agents.
Salesforce is preparing for enterprise software to shift from fixed applications toward systems where AI agents carry out work.
An AI agent, by contrast, carries out tasks within that system.
A worked example used a satirical clinic application for overworked AI agents.
OpenAI reports that it used 10,000 automated AI agents working concurrently to develop the result.
That figure describes this particular helper and test set, not the failure rate of AI agents generally.