AI agents and agent tools, sorted by the job

Agent builders, open-source frameworks, coding, research, sales, support, voice and desktop agents, plus the sandboxes, connectors and tracing tools they run on. Each tool is described in our own words, with a link to its official site.

  • Descriptions written by hand
  • No star ratings or rankings
  • Sponsored spots always labeled

How to choose an AI agent tool

68
AI agent tools
10
categories
29
are open source
10
full write-ups

Browse AI agent tools by category

Featured AI agent tools

Sponsoredhome This position is open Put your tool at the top of the home page. Paid spots are always labeled Sponsored. From $199 a month. Get featured

What a featured listing costs · from $199 a month

Tools with a full write-up

Longer, plain descriptions of widely used AI agent tools. Not a ranking, and nobody paid for these.

Open-source agent frameworksLangChain

LangGraph

LangGraph is an open-source library from the LangChain team for building agents as graphs. Each node is a step, such as a model call or a tool call, and edges…

  • Open source
  • Can run on your own servers
Read the write-up
Open-source agent frameworks

CrewAI

CrewAI is an open-source Python framework for running several agents together. You define each agent with a role, a goal and a short backstory, give it tools, and assign tasks.…

  • Open source
  • Can run on your own servers
Read the write-up
Open-source agent frameworksMicrosoft

AutoGen

AutoGen is an open-source framework from Microsoft for building applications in which several agents talk to each other to finish a task. One agent might write code, another run it,…

  • Open source
  • Can run on your own servers
Read the write-up
Coding agentsCognition

Devin

Devin is a coding agent from Cognition. You give it a task, such as fixing a bug, adding a small feature or upgrading a dependency, and it works in its…

Read the write-up
Coding agentsAll Hands AI

OpenHands

OpenHands is an open-source platform for software development agents, maintained by All Hands AI and first released under the name OpenDevin. An OpenHands agent can read and edit files, run…

  • Open source
  • Can run on your own servers
Read the write-up
Voice agents

Vapi

Vapi is a developer platform for building voice agents that talk to people over the phone or in a web app. It handles the parts of a live call that…

  • Developer API
Read the write-up

Every category, A to Z

How tools are listed

  1. 1

    Find the tool

    Tools come from submissions and from our own reading. A tool must be live and available today.

  2. 2

    Check the official site

    We open the official website, docs or code repository and confirm the tool does what our listing will say.

  3. 3

    Write it ourselves

    Each description is a short account, in our own words, of what the tool does and who it suits.

  4. 4

    Mark the gaps

    If we can't confirm a detail, such as an open-source license or MCP support, the listing says "not enough info".

A tool is listed once we have confirmed that its official site or repository is live and that it does what our description says. Each tool sits in the one category that fits its main job, even when it covers several. We don't publish scores, rankings or prices, because prices change often and a score would be our opinion presented as a fact. Tools we couldn't confirm stay off the site until we can.

A listing is not a safety review. Agents act on your behalf in other systems, so check each vendor's security documentation, permissions model and data terms yourself before you connect one to real accounts.

Makers can submit or claim a listing. A person checks every request. Published by Zwings LLC.

How to choose an AI agent tool

Start with one task, written down in a sentence or two. "Answer order status questions in chat" or "turn failing tests into pull requests" is a task. "Use agents across the business" is not. A clear task tells you which category to look in, which systems the agent has to reach and how you will know whether it worked.

Then decide how much you want to build. Agent builder platforms and finished agents for sales, support or voice get you running quickly, with the vendor deciding much of how the agent behaves. Open-source frameworks give your developers control over every step, at the cost of hosting, maintenance and more of their time. Many teams try a finished product first and move to a framework only for the work that needs custom behavior.

List every system the agent must read from or write to, and the account it will use for each. This is the most important part of the decision. An agent that can only read is low risk. One that can send email, change records, run code or spend money needs limits: narrow permissions, a cap on how many actions it can take, and a person approving the actions that are hard to undo.

Ask each vendor how a person stays in the loop. Can you require approval for chosen actions? Can staff take over a conversation or a run partway through? Can you pause the agent instantly? A good answer includes screenshots or a short demo, not a promise.

Ask to see the run history. Every agent will make mistakes, and the question is whether you can find out why. Look for a record of each step with the tool inputs and outputs, and check how long it is kept and who can read it, since it may contain customer data.

Think about cost per completed task, not cost per seat. Agents can loop, retry and call many tools, and each step may use model calls you pay for. Ask how usage is counted and whether you can set limits per agent or per run.

Run a pilot with real work and a person checking every output. A few weeks on one task, with a weekly read through the run history, will tell you more than any comparison table. Expand the agent's reach only after the record shows it handles the common cases well and fails in ways you can catch.

Frequently asked questions

What counts as an AI agent here?

We use the word for software that uses a language model to decide on steps and take actions with tools, such as searching, calling an API, editing code or sending a message, rather than only replying in a chat. We also list the frameworks, infrastructure and tracing tools used to build agents.

Does WithAutobot recommend specific tools?

No. We describe what each tool does and who it suits, in our own words, and link to the official site. There are no scores or rankings. Paid placements are marked Sponsored and do not change our descriptions.

Are AI agents safe to connect to company systems?

They can be, with care. Give each agent only the access its task needs, require human approval for actions that are hard to undo, keep a run history, and test with low-risk work first. A listing here is not a security review.

Why don't you list prices?

Agent tools are often priced by usage, seats, conversations or minutes, and pricing changes often. An out-of-date price is worse than none, so each tool page links to the official site where current pricing lives.

How do I suggest a tool or fix a listing?

Use the submission form, or email us. We check every request by hand, write the description ourselves, and correct listings when a tool changes, is renamed or shuts down.

What does "not enough info" mean on a tool page?

It means we could not confirm that detail, such as an open-source license, a self-hosted option or MCP support, from the tool's own site or docs. We leave it marked that way instead of guessing, and update it when the maker confirms.