Overview

Tasklet is an AI command center that enables you run real parts of your business on AI agents. Unlike a chatbot that just answers questions, a Tasklet agent takes ownership of work: it sends emails, updates spreadsheets, files tickets, builds apps, and runs entire business processes across your tools. You just describe your goal in plain English and Tasklet figures out the rest.

It connects to your whole stack: thousands of apps are one click away, plus any HTTP API or MCP server—even the private tools your company built itself. Agents get their own secure computers in the cloud to run code, process data, and open a browser when an app has no API. And because agents run in the cloud, they work 24x7—on a schedule or when events happen, like a new email or Slack message—while you decide what they can do on their own and what needs your approval first.

Tasklet is built for business teams. You can share agents, connections, and knowledge across the company, with credentials locked down and access set once. You get the best models from Anthropic, OpenAI, Google, and open-source labs in a single plan—routed to whichever does each job best—plus granular access controls, centralized billing, and SOC 2–compliant security.

Agents

The core building block in Tasklet is the Agent: a persistent AI worker that learns and improves as you work with it over time. You grant agents access to Connections and Knowledge so they can work across your services and to your standards. Automations let them work in the background, even when you’re offline and your computer is closed.

You can create different agents for different areas of work. For example, you might create a recruiting agent that knows your hiring process and can access your applicant tracking system, or a support agent that knows your policies and can work with your help desk.

Create an agent

Select New agent in the agent rail. The setup flow helps you define the agent’s role, describe the work you want it to do, choose the Connections it needs, and give it starting instructions.

Threads

Each thread is a focused conversation with an agent. Create a new thread for each task, project, or piece of work, rather than putting unrelated work into one long conversation.

Threads remain separate and focused, while drawing from the same agent’s instructions, Knowledge, Connections, and Memory. For example, a marketing agent could have one thread for each launch, another for reviewing campaign performance, and another for drafting customer emails.

Memory

Memory is what an agent learns from relevant work across all its threads. It includes the conversations themselves, along with the files, images, and attachments the agent worked with. What the agent learns in one thread can help it in another, and it can search earlier threads when it needs more context.

Memory updates asynchronously in the background, so there is nothing to manage day to day. New agents use Memory automatically. To turn it on for an existing agent, open the agent’s sidebar and select More → Advanced, then find Agent-wide memory and select Turn on. Once Memory is enabled for an agent, it cannot be turned off. Your other agents are not affected.

Sharing an agent

Agents are private when you create them, but you can invite teammates when you want to work together. See Shared Agents to learn what members can access and how to share an agent.

💡 Tip: Think of each agent as a boundary for access and learning. Create a new agent when work should be kept separate—for example, for a different team, customer, or sensitivity level—or when you want a distinct role or personality. Otherwise, start a new thread with an existing agent.

Connections

Connections let agents work with external services and accounts. Tasklet has over 3,000 built-in integrations, and it can connect to any HTTP API—including private APIs—or MCP server. It can also access services through a browser when no integration or API is available.

To connect a service, ask “Connect my Gmail” or open Settings → Connections.

How Connections Work

1. Connect your account — Tasklet opens a secure authorization flow. You sign in directly with the service.
2. Add the Connection to an agent — Open the agent’s Connections tab and select Add connection. A Connection can be added to more than one agent.
3. Choose which tools the agent can use — Each Connection provides tools such as reading email or creating a task. Tool permissions are configured separately for each agent, so different agents can receive different levels of access to the same Connection.

When an agent needs a tool that has not been enabled, it asks for permission and shows you a card to approve the request.

💡 Tip: You can ask an agent to connect a service, add a Connection, or show you the tools available for a service.

What you can connect

Here are some of the most popular built-in connections by category:

Custom APIs & Integrations

Tasklet can connect to any HTTP API or MCP server directly. Describe what you want to connect to, and Tasklet will set up a connection flow where you can securely provide your credentials.

“Connect me to our inventory API so I can monitor stock levels and alert team members when we’re running low.”

You can also connect to internal company APIs. Share instructions or reference files with Tasklet and it will learn to make requests.

Security protections

• Your credentials are encrypted and stored securely within the Connection flow—Tasklet never sees your passwords.
• Connections use OAuth when available, so you sign in directly with the service.
• You control which tools each agent can use with a Connection.
• You can manage and revoke your Connections anytime from Settings → Connections.

Sharing a connection

Connections are private by default. You can share one when teammates should be able to add it to their own agents without seeing its credentials. See Shared Connections for sharing roles and instructions.

Automations

Create an automation by asking Tasklet what should happen and when. Automations can run on a schedule, when emails arrive, or in response to external events.

How to create an automation

Describe what you want to happen and when:

“Send me a daily briefing every morning at 9am.”

“When I get an email with an attachment, file it in Drive.”

You can also say “create an automation for…” to be explicit.

Types of automations

💡 Tip: You can add a filter to Gmail automations to reduce the number of emails being processed and lower credit usage. For example, you can filter by attachments only, specific senders, or emails marked as important. To do this, specify the filter when creating the automation: “When I get an email with an attachment…”

To test, delete, or view details of an automation, ask in the thread or select the automation icon (⚡) in the thread header.

Multiple automations per thread

A single thread can handle multiple automations. For example, an inbox assistant thread could:

• Run a daily briefing every morning
• Draft replies for incoming emails
• Star important messages automatically

Text & Email Messaging

Tasklet can send and receive messages via email or text (iMessage & RCS). When someone responds, your thread receives the reply and can take action.

Email anyone (with permission)

You can add any email address as a contact method. The recipient clicks a verification link to opt in, and then your thread can send them messages—reports, notifications, requests for approval, etc.

Text messaging

Tasklet can also send and receive text messages via iMessage and RCS. This is useful for time-sensitive alerts or when you want to interact with your thread on the go.

Two-way conversations

Recipients can reply to Tasklet’s emails or texts. Your thread receives the response and can continue the conversation or take action based on it. Recipients don’t need a Tasklet account.

How to set it up

1. Ask in any thread to add a contact method (email or phone number)
2. The recipient verifies—clicking a link for email, or sending a text for phone numbers
3. Once verified, your thread can message them anytime

Example prompts

• “Every Friday, email the sales team a summary of closed deals”
• “Text me if any emails come in from my boss”
• “Include my teammate in these reports, their email is…”

Knowledge

Knowledge is the set of Custom Instructions, Skills, and Drives an agent can use across all its Threads.

Custom Instructions

Custom Instructions are house rules the agent follows in every Thread—tone of voice, formatting, things to always or never do, and basic context about your company. Set them once instead of repeating yourself in every conversation.

Examples of useful instructions:

• “Always write in British English and use our brand name ‘Acme’ (never ‘ACME’).”
• “Our fiscal year starts in February—assume that for any reporting.”

💡 Tip: Instructions are capped at 10,000 characters. Keep them focused on rules and context, and put longer reference material in files instead.

Skills

Skills are reusable, step-by-step playbooks for work your agent does repeatedly. They help the agent perform the same kind of task consistently.

Each Skill has a name, a short description of when to use it, and its instructions, plus any supporting files. Threads see all available Skills and reach for the right one when a task matches.

You create and manage Skills from the same Knowledge area:

• Pre-built Skills — install ready-made Skills from the gallery with one click.
• Custom Skills — create your own: give it a name, a brief description, and the instructions the thread should follow.

Examples of Skills you might create:

• A weekly report Skill that defines exactly how your reports should be structured and sourced.
• A support triage Skill that encodes how to categorize and route incoming tickets.

Drives

Drives organize the files and Skills an agent can use. Every agent has its own Drive for files and Skills specific to that agent. The agent can create and update its contents as it works, and they remain available across all of the agent’s Threads.

You can also give agents access to Shared Drives, which allows multiple agents to reference and modify a shared set of files.

The agent uses other files in its Drives when they are relevant:

• “What’s our refund policy?” → the agent reads it from a policy document in one of its Drives.
• “Draft a proposal using our standard template.” → the agent starts from a template in one of its Drives.

💡 Tip: Not sure whether to use instructions, files, or a Skill? Use instructions for always-on rules, files for reference material, and a Skill for a repeatable procedure with clear steps.

Browser

Every Tasklet thread has a built-in web browser it can use to visit and interact with any website. The thread can use it for example to access authenticated sites or fill out forms.

How to use it

Just describe what you need and Tasklet will open a browser if it helps:

Log into my Google Ads account and take a screenshot of this week’s campaign performance dashboard.

You can also say “use the browser to…” to be explicit, but usually you should let the thread choose the best approach. In many cases using a built-in integration or non-interactive web tools will be faster, more reliable, and use fewer credits.

What it can do

• Navigate complex pages, enter text, interact with buttons and other elements
• Fill out forms and click through multi-step flows
• Extract data from pages (tables, listings, article text)
• Upload and download files
• Scroll long pages and interact with dynamic content
• Hand off sign-in or CAPTCHA steps to you when a site requires them

Logins and CAPTCHAs

Tasklet will never ask you for your password. When a site needs you to sign in, or a CAPTCHA appears, the thread navigates to the login page and shows you a live browser preview in the conversation. You take over in the preview—enter your credentials or solve the challenge directly—and then hand control back. The thread continues from there.

Saved logins and cookies

The browser keeps a persistent profile for each thread. Once you’ve signed into a site through the browser preview, the thread stays signed in for future tasks—you don’t have to re-enter your credentials each time. Cookies, site preferences, and any data the site stored locally carry over the same way.

If a thread goes unused for over 30 days, the browser’s saved state may be cleared and you’ll need to sign in again the next time the thread visits that site.

Example prompts

• “Log into my company’s Workday portal and submit my timesheet for this week.”
• “Log into my Google Ads account and take a screenshot of this week’s campaign performance dashboard.”
• “Log into Gusto and download the last 3 payroll summary PDFs.”
• “Go to the DMV appointment scheduler, find the earliest available slot in San Jose, and book it for me.”
• “Search Google Flights from SFO to NYC for next Friday, apply ‘nonstop only,’ sort by price, and screenshot the top results.”
• “Log into my accountant’s client portal and upload these three PDF documents.”
• “Go to DoorDash and re-order my last Thai food order.”

Instant Apps

Instant apps are lightweight, interactive web apps that run inside Tasklet’s preview panel. Use them when you need an interface to click through, filter, sort, or enter data while continuing your conversation.

Instant apps are a good fit for dashboards you want to drill into, trackers or lists with inline edits, interactive calculators or forms, kanban-style workflows, and plenty more.

What makes instant apps powerful:

• Live data — Apps can use your connected services to show current data, not mockups.
• Two-way sync — Changes you make in the app write back to your tools.
• Feedback loop — Actions you take in the app flow back to the thread, so it can respond to what you clicked or edited.
• Saved — Apps are saved to the thread’s filesystem, so you can revisit or update them later.

Example prompts:

• “Build a dashboard showing my open Linear issues grouped by priority.”
• “Create a kanban board for the projects in my Notion database.”
• “Make an interactive chart of monthly revenue from the sales spreadsheet.”

Tasklet for Mac

The Tasklet desktop app gives you the full Tasklet experience with native Mac features. It supports Apple silicon Macs running macOS 13 or later.

Install the desktop app

1. Get the desktop app.
2. Open the downloaded image file (.dmg).
3. Drag Tasklet into the Applications folder.
4. Open Tasklet and sign in to your account.

Tasklet checks for updates automatically. To check manually, select Tasklet → Check for Updates… in the menu bar.

Start work with Quick Chat

Quick Chat is currently in preview. Press ⌥-Space from any Mac app to open it. Choose an agent and an existing thread, or start a new thread. You can type or dictate a message, attach files, choose a model, and send without first switching to the main Tasklet window.

After you send, it will open Tasklet to the chat. Press ⌥-Space again or click outside Quick Chat to close it without sending.

To turn off Quick Chat, disable Quick Chat in Experiments settings.

Use tabs, windows, and shortcuts

The desktop app supports native tabs and multiple windows. Use these shortcuts to move faster:

• Command-T: Open a new tab
• Command-N: Open a new window
• Control-1 through Control-9: Switch agents
• Command-Shift-A: Archive the current thread

Right-click a link to open it in a new tab or in your default browser.

See keyboard shortcuts to see the complete list.

Add an Instant App to your Dock

You can install an Instant App as an independent Mac app. Open the app preview, select its actions menu, then select Install to Mac. Tasklet adds the app to the Tasklet Apps folder in your user Applications folder and opens it in the Dock with its own name and icon.

Open the installed app from the Dock to return to that Instant App. To remove it, open the app preview in Tasklet and select Uninstall from Mac.

Teams & Sharing

Tasklet lets your team share the resources agents need and collaborate in shared agents. Invite teammates to your organization, give them access to Connections and Drives they can use with their own agents, or add them to an agent when you want to work together in the same threads.

Organizations

An organization is your team’s home in Tasklet. It groups your members, agents, and billing in one place.

How Tasklet is organized

• Organization — your company’s Tasklet account. It contains your members and agents and manages billing.
• Members — the people in your organization. Each person has an organization role: Owner, Admin, or Member.
• Agents — persistent AI workers, each with a role and access to the Connections and Knowledge they need.

Organization roles

Inviting people

People join an organization by invitation. By default, any member can invite someone. An Admin can restrict invitations to Admins only from Settings → Organization → Members.

When inviting someone, choose whether they should join as a Member or Admin. Their role can be changed later from the same page.

Working as a Team

Tasklet supports two ways to work as a team:

Work together in shared agents

A shared agent works more like a teammate who collaborates closely with a group. Its members work in the same threads and share the agent’s context and resources.

This works well when the work itself should be collaborative—for example, a recruiting agent used by a hiring team or a support agent used by everyone handling customer requests.

Use private agents with shared resources

A private agent works primarily for you, like an executive assistant. Teammates can share Connections and Drives with you, then you can add those resources to your own agents.

This works well when people want their own agents and threads but need access to the same company accounts and information.

What you can share

Shared Agents

Share an agent when several people should collaborate on the same work. Members can see its threads, continue work started by someone else, and use the context and resources available to the agent.

Because the agent’s work is shared, every member has visibility into its threads. This makes shared agents useful for ongoing team workflows where context, activity, and results should stay together.

What agent members can access

When someone becomes a member of an agent, they can:

• See and use every thread in the agent.
• Start new threads and continue work started by another member.
• Access the agent’s Knowledge.
• Use connections added to the agent through the tools enabled for that agent.

A connection inside a shared agent is available to all of the agent’s members through that agent. This does not give those members direct access to add the connection to their other agents. To do that, the connection owner would share the connection with them separately.

💡 Tip: Give someone access through a shared agent when you want the work they do with a connection to remain visible in that agent. Share the connection directly when they should be able to use it privately with their own agents.

All agent members currently have equal access. Agent members can add or remove other members, start threads, rename the agent, and archive it. Organization Owners and Admins can also manage access to any agent in the organization.

Share an agent

1. Open the agent you want to share.
2. Click the share icon in the sidebar.
3. Add an existing member of your organization.

💡 Tip: If the person is not yet in your organization, you can invite them while sharing the agent. They’ll receive an invitation to join your organization and access the agent.

You can also see all agents you have access to from Settings → Agents.

Shared Connections

Connections are private by default. You can share a connection with other members of your team so their agents can work with a service or account you have already connected.

Common types of connections to share include:

• Internal systems — company APIs and MCP servers
• Service accounts — dedicated credentials used for billing, analytics, and other shared operations
• Shared accounts — company-managed accounts such as a support inbox, social media account, ecommerce store, or advertising account

Connections you might keep private include your personal email, calendar, or accounts that only you should have access to. Sharing a connection never reveals its API keys, login details, or other underlying credentials.

Share a connection

1. Open Settings → Connections.
2. Select the connection you want to share.
3. Choose the people—or your entire organization—who should receive access.
4. Choose their access role:

💡 Tip: If the person is not yet in your organization, you can invite them while sharing the connection. They’ll receive an invitation to join your organization and access the connection.

Use a connection someone shared with you

1. Open the agent where you want to use it.
2. Open the agent’s Connections tab.
3. Click Add connection.
4. Click Add on any shared connection to add it to the current agent.
5. Choose which tools the connection can use in that agent.

You can also see all connections you have access to from Settings → Connections.

💡 Tip: You can ask an agent to add or share a connection, and it will show you a card to complete the action. You can also ask it to use a tool it does not yet have permission to use; it will show you a card to approve the request.

Shared Drives

Every agent has its own private Drive. Shared Drives allow agents across your organization to reference and modify the same set of files.

Shared Drives are like Google Drive for the agents across your organization. Multiple agents can collaborate using the same files and use centralized Skills. Agents are automatically aware of Skills in Shared Drives they can access and automatically receive updates to those Skills.

People can also view and edit content in Shared Drives directly.

Use Shared Drives to:

• Share a library of approved Skills across multiple agents.
• Let several agents create or update the same project files.
• Keep sensitive material limited to specific people and agents.
• Separate team, customer, or workflow-specific context from company-wide knowledge.

Organization Drive

The Organization Drive is a special Shared Drive created automatically for every organization. It is added to every existing and new agent, so it is the simplest place to put Skills and files that should be available company-wide.

For example, you might put your company definitions, brand guidance, policies, and standard operating procedures in the Organization Drive.

💡 Tip: You can ask your agents to review the context and Skills they have and recommend which things would be beneficial to move to the shared Organization Drive.

Create a Shared Drive

You can create additional Shared Drives when you need select content to have different access boundaries than your Organization Drive—for example, a Finance Drive, a customer project Drive, or separate libraries of locked and editable Skills.

To create one:

1. Open Settings → Drives.
2. Under Create new Drive, enter a display name and mount path, then select OK.
3. Share it with the people who should have access.
4. Add it to each agent that should use it.

💡 Tip: You can also ask the agent to create a new Shared Drive. This will create the Drive and add it to the current agent.

Add a Shared Drive

Only the Organization Drive is automatically added to agents. For additional Shared Drives, you must add these to the agents that need them:

1. Open the agent you want to use it.
2. Open the agent’s Knowledge tab.
3. Select Add Drives.
4. Select a Shared Drive, choose View or Edit access for the agent, then select Add drive.

Access Levels

You can manage what access people and agents have to Shared Drives:

• View — See and read the Drive’s contents.
• Edit — Add, change, move, or delete files in the Drive.
• Manage — Edit the Drive and control which people have access. Manage access is available only to people.

People can grant an agent up to the same level of access they have. For example, someone with View access can give an agent View access, while someone with Edit access can give an agent Edit access.

Give an agent View access when it should use files or Skills without changing them. Give it Edit access when it should create, maintain, or improve the shared contents as it works.

Sharing a Drive with a person does not automatically add that Drive to their agents.

Additional Capabilities

File Processing

Tasklet can transfer files between connections (Gmail, Drive, Slack, APIs, etc.), process them, and create new files.

What it can do

Combine files with automations

You can combine file processing with automations to run recurring workflows:

• Every morning, pull data from your CRM and generate a summary report.
• When receipts arrive in your inbox, extract the data and add a row to a spreadsheet.
• On a schedule, check a public dataset and alert you if anything changes.

Example prompts

• “When I get an invoice emailed to me, save it to my ‘Invoices/2026’ folder in Drive and log it in my expenses spreadsheet.”
• “Analyze this CSV and send me a summary with charts.”
• “Resize all images in this folder to thumbnails.”
• “Download all my receipts from Doordash and upload them to Drive.”

Media

Tasklet can create images and short videos directly in the app. It can also transcribe audio and video files so you can summarize, analyze, or use their contents in other work.

What it can do

Example prompts

• “Create a product image with a transparent background.”
• “Generate a short launch video for this new feature.”
• “Transcribe this customer call and pull out the main requests.”

Sandbox (Code Execution)

Every Tasklet thread has a Linux sandbox for running code and processing files.

Pre-installed tools:

• ffmpeg — audio and video processing
• ImageMagick — image manipulation and conversion
• Pandoc — document format conversion
• curl, jq — web requests and JSON processing
• Python 3.12 with access to any package (pandas, numpy, matplotlib, etc.)

Installing packages:

The sandbox can install any additional package on demand. If a task requires a specific library or tool, Tasklet installs it and proceeds.

Filesystem access:

Code running in the sandbox can read and write to the thread’s filesystem. Scripts can be saved and rerun later, or process files that have accumulated over multiple sessions.

Example prompts:

• “Analyze this 10,000-row CSV and give me summary statistics with visualizations.”
• “Convert all these Word docs to PDFs.”
• “Write a script that checks this API every hour and alerts me if the price drops.”

SQL Database

Every thread has its own persistent SQL database to store and retrieve structured data across sessions.

What it’s for:

• Tracking progress on recurring tasks
• Avoiding duplicate work (“have I already processed this?”)
• Storing information that builds up over time
• Keeping configuration that changes between runs

If you want your thread to remember something specific, you can tell it explicitly:

• “Store all the contacts you find in your database so you don’t look them up again.”
• “Remember everything we learn about Acme Corp in your database.”
• “Track all the vendors we’ve evaluated, their pricing, and our notes in the database.”

Subagents

For complex tasks, Tasklet spins up focused subagents to handle specific parts of the work, like mini-workers with clear instructions for their job.

How they work:

Subagents are each focused on a specific task. When finished, they report back the result, keeping work organized and costs down.

Why it matters:

• Breaks big tasks into smaller, manageable steps
• Keeps costs down: each subagent works with only what it needs
• Processes long lists without losing track (e.g., reviewing 50 resumes one by one)
• Handles heavy lifting like web scraping or document analysis in isolation

Tasklet often creates subagents automatically, but you can request them explicitly:

• “Create a subagent to research each competitor, then compile the results into one report.”
• “Use a subagent to review each resume in this folder and score it against our job criteria.”
• “For each lead in this list, have a subagent research the company and draft a personalized outreach email.”

Webhooks (External API)

Tasklet can receive data from external services through webhooks. Each webhook provides an API endpoint that starts an AI-powered workflow.

How to set it up:

“Set up a webhook automation that runs when I get a new order from my Shopify store.”

Tasklet creates a unique URL. Point your external service at that URL, and when it sends data, Tasklet wakes up and processes it.

Examples:

• Stripe sends a webhook when a payment succeeds → Tasklet updates your CRM and sends a welcome email
• GitHub sends a webhook on new issues → Tasklet triages and labels them
• Your app sends a webhook when users sign up → Tasklet enriches profiles and notifies sales
• Meeting transcription service sends a webhook → Tasklet creates tasks and emails a recap

Thread-to-thread communication:

Webhooks also enable threads to talk to each other. One thread can make a Direct API call to another thread’s webhook URL to start it. This is useful for coordinating work between specialized threads.