Picture this: You walk into a newly opened restaurant for the first time, excited by the positive reviews, and confidently stride to a window-side table. As soon as you're settled, the waiter approaches, but before they can speak, you say, "The usual, please."
The waiter stares at you like you've lost your mind. They've never seen you before. They have no idea what "the usual" means.
That's your AI agent every time you start a new conversation. It has absolutely no idea who you are, what you want, or why you're asking it to "check my tasks." Which tasks? Whose tasks? Who even are you?
This same awkward interaction happens with many AI agents, especially those with no recollection of previous interactions. Each time you start a conversation, they're essentially "meeting" you for the first time, requiring either you or the agent developer to provide a whole backstory about your identity and preferences before they can actually help.
For developers building AI agents, this means every user interaction starts with a painful game of 20 questions. Your users have to explain their entire identity before the agent can do anything useful.
We Know Who You Are!
Think about how many times this happens with simple requests:
- "Book a meeting with Sarah" (Which calendar? What timezone are you in?)
- "Update my status in Slack" (Which workspace? What's your user ID?)
- "List my tasks on my team space" (Which team? What's your workspace?)
Without the necessary context upfront, the agent doesn't know who you are or which workspace is yours. It has to start a round of basic questions before actually doing anything useful. For such simple requests, having to do this each time is tiresome and annoying. You could solve this by adding end-user specific memory to your application/agent, but the same problem remains - how do you know when to load the relevant memory for the right services or use cases? Furthermore… Why ask your users for information the bot should already, theoretically, know?
Aiming to improve our customers' experiences when using our tools in their agents, we at Arcade realized the importance of offering LLM Agent builders a simple and standardized way to fetch basic user information when needed. That's why we created a new type of tool we call "WhoAmI" tools. Their goal is simple: introduce you to the Agent when needed.
WhoAmI tools are like a digital ID card for your agent interactions. They tell the agent exactly who you are in each service – your Slack handle, your Jira workspace, your Gmail address – without making you spell it out every damn time. The information a WhoAmI tool retrieves depends on what the provider makes available: user unique identifier, name, contact information, profile picture, timezone, and more.
There Are Multiple You's!
Here's the thing: the question Who am I? isn't as straightforward as it seems. Who you are depends on your current context. At home you're a parent, at work you're a boss, at your favorite MMORPG you're CutePieKitty1337. Just as you wear many hats in the physical world, so it is in the digital one.
At work you're john.smith@company.com. On Discord you're xXDragonSlayer420Xx. On LinkedIn you're 'Senior Solutions Architect.' Each service knows a different version of you, and your AI agent needs to juggle all these identities without mixing them up.
You are a different you on each one of your profiles across different providers. They each contain a set of information that composes your persona there – some set by you as preferences, others set automatically like IDs. Some of this info matches between different providers and some doesn't.
To have proper context of who you are at a specific provider, you need selected pieces of the data they have about you. Arcade's WhoAmI tools were built to fetch this information for your agent, as long as you give it the required authorization.
Show Me It Works!
Now let's see it working with our fresh-baked ClickUp toolkit, which has had the WhoAmI tool since day one:
This shows the interaction with a ClickUp chat agent for "the first time" where my first request is to get tasks assigned to me. Even though the agent has some basic greeting context, it has no idea of Who I am in ClickUp. It needs my ClickUp user_id and workspace_id.
But here's the magic: instead of asking me twenty questions, the agent proactively calls the Clickup_WhoAmI tool, gets the info it needs, then immediately lists my tasks. No awkward back-and-forth. No "what's your workspace ID?" Just results.
Even when the agent initially fails to comprehend, it can utilize the WhoAmI tool to acquire the necessary information for further tool calls. Arcade's built-in error recovery features, like RetryableToolErrors, provide crucial feedback to the agent with LLM-tailored error descriptions and recovery guidance. This helps it understand that invoking the WhoAmI tool would solve the problem. This iterative process allows the agent to retry the call effectively – and achieving this functionality is way easier than you'd think!
Try It Yourself!
At Arcade we're always looking for new approaches to improve our AI toolkits and offer developers the best experience possible when building Arcade-powered agents. Better developer experience means smoother agent interactions for end users. This is the difference between a demo that looks cool and an agent that actually ships to production.
That's why we've added WhoAmI tools to our most popular toolkits. They're totally optional – the power to use them or not is yours!
Stop making your users explain who they are. WhoAmI tools are live right now in:
Google Suite:
- Gmail
- Google Calendar
- Google Contacts
- Google Drive
- Google Docs
- Google Sheets
- Google Slides
Microsoft:
- Outlook Mail
- Outlook Calendar
- Microsoft Teams
- SharePoint
And more:
- Slack
- Jira
- Notion
- Zendesk
Your agents can finally remember who they're talking to. Try it yourself – your users will thank you for not asking them their email for the 47th time today.
Build agents that actually remember who they're talking to.
Get started with Arcade.dev →
See all WhoAmI-enabled tools →
Setup time: 6 minutes. User frustration eliminated: Priceless.
Francisco Liberal is an AI Integrations Engineer at Arcade.dev, where he builds the tools that make AI agents actually useful. A full-stack developer and .NET expert with a passion for quality and efficiency, Francisco specializes in creating robust integrations that bridge the gap between AI capabilities and real-world applications. When he's not solving identity crises for AI agents, you'll find him advocating for automated tests and reminding everyone: don't forget to be awesome!
