Today we are announcing the release of Agent Mode in Warp AI, a new way to interface with AI from the command line so developers can accomplish multi-step workflows without leaving the terminal.
Agent Mode can:
Agent Mode raises the level of abstraction in the terminal. Using natural language, you can ask the terminal to accomplish any high level task without worrying about the specific commands you need.
Agent Mode knows when it needs more information to help you complete a task. It will ask permission to run commands on your machine and use the outputs to guide you, step-by-step. It’s also self-correcting when it runs into errors.
Because Agent Mode executes tasks through the CLI, it integrates with practically any service with zero configuration. If the service has a CLI, an API, or publicly available docs, you can use Agent Mode for the task. Agent Mode has inherent knowledge of most public CLIs, and you can easily teach it how to use internal CLIs by asking it to read their help content.
Watch Agent Mode in action:
Browse more examples and learn how it works.
Agent Mode makes it easy (and even fun!) to work through development tasks that would otherwise require a lot of time and context switching.
For example, if you would previously browse through the AWS console to figure out how to provision a new instance or upgrade your database, with Agent Mode you can type “help me upgrade an AWS database” and get guidance without leaving your terminal. Provide a little context (e.g. the region, the service) and Agent Mode will walk you through the process step by step, gathering info and fixing errors until you have completed the task together.
Because the CLI is already a universal interface, Agent Mode is automatically “integrated” with almost any dev tool out of the box. You can use Agent Mode to interface with Github (gh), AWS/GCP (aws/gcp), kubernetes, Datadog (dog), and any other tool with a CLI. You can use curl to read web pages and interact with APIs. It can even interact with internal CLI tools. As long as the tool has a --help option, you can ask Agent Mode to learn it, and then immediately have Agent Mode start doing tasks with it. You can also attach specific output or errors for the AI to reference.
Much more than codegen or a chatbot, you can use Agent Mode for use cases like debugging, devops and analysis.
Let’s start with the simplest example: Say you’ve run into the super-common “port 3000 already taken” error. You can type “fix it,” attach the error context, and Agent Mode will respond to help.
Notably, the AI agent is self-correcting. When it makes a mistake, it adjusts itself until it completes the task for you. For example, if Agent Mode suggests a command with an incorrect flag, it will follow up with another command with the right flag. Agent Mode will make sure commands work and are suited to your system.
When you have auto-detection enabled for natural language, Warp AI will automatically detect when you are entering plain English on the command line and switch into Agent Mode.
Auto-detection happens locally. The Warp app has a local classifier that checks input strings. Nothing you type in the input ever leaves your machine during the natural language detection classification. After auto-detection occurs, you must take explicit action to send a request to Warp AI. You can disable auto-detection at any time.
You have total, explicit, and granular control over any information that is sent to Warp AI. You get to opt-in specific terminal outputs (if any) to send to the AI.
Once you are engaging with Agent Mode, Warp AI will read the command outputs for any command you authorize it to run during the session as it gathers information in pursuit of completing a task. In other words, if Agent Mode requests that you run a command on its behalf, it will read the output of that command. You are always in the driver’s seat to approve each command it runs.
Agent Mode’s approach to context-gathering is transparent since it is all done through terminal commands. For example, if Warp AI needs the name of your git branch, it will ask to run git branch and read the output. You will be able to approve or reject the command. We’ve designed Agent Mode so that you know exactly what information is leaving your machine.
Since Warp AI uses OpenAI, OpenAI’s servers will receive all input. OpenAI does not train their models on this data. If your organization requires Zero Data Retention or the option to bring your own LLM, these features are available on Warp’s Enterprise plan. Contact us to learn more.
Existing Warp users will notice that new Agent Mode has replaced the first-generation Warp AI chat panel.
Agent Mode can handle all of your Warp AI chat use cases and more. It’s much more powerful and does not require you to copy and paste as you go.
AI Command Suggestions (press “#” to ask Warp AI) will remain supported for the short-term. You may find this helpful as a way to get quick suggestions without engaging Agent Mode in a session. As we confirm that Agent Mode can deliver accurate suggestions for commands with equal speed and success, it’s likely we’ll sunset this feature in favor of the Agent Mode experience.
Agent Mode is available today for Warp users on every plan tier. The Free plan includes up to 40 AI requests per month and higher requests are available on Warp’s Pro, Team, and Enterprise plans. Compare plans.
To learn more about Agent Mode:
Give it a try and let us know what you think! We’d love for you to share your first impressions and use cases with us on X (Twitter) @warpdotdev.