A Practical Intro to AI Agents and Agentic AI

If you want to learn:

- How to build a multi-tool AI agent across Gmail Calendar Slack and Notion

- How AI agents can prepare you for meetings using context from multiple apps

- Why combining Calendar Slack and Notion makes AI agents far more useful

- How Claude Cowork and OpenWork orchestrate multi-tool workflow automation

- How to save 30 minutes per meeting with AI-driven prep briefs

Then this lecture is for you!

Move beyond single-purpose email agents and build a multi-tool AI agent that pulls context from Gmail Calendar Slack Notion and the open web inside Claude Cowork and OpenWork. The lecture frames the workday problem: real work spans inbox calendar knowledge base and team chat, and information gets lost across silos. You will set up an AI agent that produces a clean one-page meeting brief with company overview, relationship history, project context and three talking points so you walk into any external call ready in seconds. This is the foundation for the rest of the section on multi-tool workflow automation.

If you want to learn:

- How to connect Google Calendar Notion and Slack to Claude Cowork

- How to add multiple connectors to a Claude Cowork AI agent in minutes

- What read and write tools each Claude Cowork connector exposes

- Why Calendar Notion and Slack are the core stack for an AI productivity agent

- How Claude Cowork Customize and Browse Connectors work for workflow automation

Then this lecture is for you!

Connect Google Calendar Notion and Slack to your Claude Cowork AI agent through the native one-click connectors so a single agent can read your calendar, manage Notion pages, and send Slack messages. The lecture walks through Customize and Browse Connectors, adds each integration with OAuth, and confirms the read-only, write, and delete tools each one exposes — including drafting Slack messages, creating Notion pages, and managing calendar events. The result is a Claude Cowork agent wired into four apps with Gmail already included, ready for multi-tool workflow automation across your daily productivity stack.

If you want to learn:

- How to test Google Calendar Notion and Slack connectors in Claude Cowork

- How a Claude Cowork AI agent reasons across multiple data sources at once

- How to query calendar events Notion pages and Slack channels from one prompt

- How AI agents combine context from different apps in a single workflow

- Why testing connectors before building agents prevents broken automations

Then this lecture is for you!

Verify that your Google Calendar Notion and Slack connectors actually work inside Claude Cowork by running multi-tool queries against a real project. The lecture asks the AI agent for today's calendar events, all Notion pages tied to a Meridian Finance project, and the related Slack channels, then shows how Cloud Cowork enriches its working context with each new tool and reasons across them in one task. You will see how a Claude Cowork agent stitches Calendar plus Notion plus Slack into a single response and why this multi-source context is the foundation for serious AI workflow automation.

If you want to learn:

- How to connect Notion to OpenWork using its built-in one-click MCP connector

- How OpenWork exposes a Notion search tool to your AI agent

- How to verify a Notion MCP integration is working with a simple query

- Why some MCP servers ship with OpenWork while others need manual setup

- How an open-source AI agent reads and reasons over Notion pages

Then this lecture is for you!

Connect Notion to OpenWork through the built-in one-click MCP connector so your open-source AI agent can read and reason over your Notion workspace without writing any code. The lecture walks through OpenWork settings and extensions, authorizes the Notion integration with a single click, and tests it by asking the AI agent to find all Notion pages tied to a Meridian Finance project. You will see the Notion search MCP tool fire from inside OpenWork and watch the agent pull page contents back through the Model Context Protocol — proof that OpenWork can match Claude Cowork on the Notion side without paid tooling.

If you want to learn:

- How to connect Google Calendar to OpenWork using a custom MCP server

- How to add Google Calendar API to an existing Google Cloud project

- How to install an open-source Google Calendar MCP via NPX

- How to edit the OpenWork config JSON to register a new MCP extension

- How to fix common credential and username errors during MCP setup

- How OpenWork uses Model Context Protocol to query your calendar events

Then this lecture is for you!

Connect Google Calendar to OpenWork using a custom open-source Model Context Protocol MCP server that reuses the Google Cloud project and OAuth credentials from the Gmail setup. The lecture enables the Google Calendar API in console.cloud.google.com, installs the calendar MCP via NPX, exports the OAuth credentials variable, edits the OpenWork JSON config file to register the new extension, and debugs a common username path error. You will then prove the connection works by asking your OpenWork AI agent for today's calendar events and matching the result against your Google Calendar.

If you want to learn:

- How to connect Slack to OpenWork through a custom MCP server

- How to create a Slack bot at api.slack.com from scratch

- Which OAuth scopes a Slack AI agent bot needs to read and post messages

- How to register a Slack MCP in the OpenWork config JSON with a bot token and team ID

- How to invite a Slack bot into channels so OpenWork can read messages

- How to test the Slack integration by querying a project channel from your AI agent

Then this lecture is for you!

Wire Slack into your OpenWork AI agent through a custom Model Context Protocol MCP server with a brand-new Slack bot. The lecture creates an OpenWork bot at api.slack.com, configures the OAuth scopes for channels history, channels read, chat write, reactions write, users read, and groups history, installs it into the workspace, captures the bot token and team ID, and registers the Slack MCP in the OpenWork config JSON. You will then invite the bot into Project Phoenix and ask your OpenWork AI agent to summarize the latest Slack discussion through the new MCP tool.

If you want to learn:

- How Claude Cowork and OpenWork compare on real Slack workflow automation

- How to build a Slack decision timeline AI agent that summarizes project channels

- How to turn long Slack conversations into structured executive summaries

- Which AI agent is faster between Claude Cowork with Opus and OpenWork with BigPickle

- How to format AI agent output as a project timeline with action items and risks

Then this lecture is for you!

Run the same Slack decision timeline task head-to-head in Claude Cowork and OpenWork to see which AI agent fits your workflow. The lecture feeds both agents the Project Phoenix Slack channel and asks for a structured decision timeline covering phases, owners, action items, open risks, and key personnel. You will compare Claude Cowork using Opus 4.7 producing a polished DOCX deliverable against OpenWork using the open-source BigPickle model returning a fast inline summary, and learn the real trade-offs between paid and free AI agent stacks for Slack summarization and project status reporting.

If you want to learn:

- How to build an AI meeting prep agent across Gmail Calendar Slack Notion and the web

- How AI agents combine internal data with external news for a one-page briefing

- How Claude Cowork plans across multiple data sources before executing

- Why OpenWork is faster but less strategic than Claude Cowork on multi-tool tasks

- How AI agents save 30 minutes per meeting on external client prep

Then this lecture is for you!

Build a real AI meeting prep agent that pulls context from Gmail Calendar Slack Notion and the open web to produce a one-pager brief: company overview, relationship history, recent external news, internal context, and three talking points. The lecture runs the same task in Claude Cowork and OpenWork against a meeting with Sarah Chen, comparing how Claude Cowork plans first with Opus 4.7 before executing while OpenWork moves faster but less strategically. You will see why multi-tool, multi-source AI agents are the real power-up for daily productivity beyond any single-app workflow automation.

If you want to learn:

- What the Model Context Protocol MCP is and why AI agents need it

- How MCP works as a USB port that unifies different software for AI agents

- Why MCP servers exist for Gmail Calendar Slack and Notion

- How MCP relates to the LLM memory tools and reasoning of an AI agent

- Why Anthropic introduced MCP and how it enables agentic AI workflows

- How custom MCP servers go beyond official APIs to extend AI agent functionality

Then this lecture is for you!

Understand the Model Context Protocol MCP — the standardized protocol Anthropic introduced that lets AI agents connect to any data source through a unified USB-style port. The lecture maps MCP back to the four AI agent architecture pillars (LLM memory tools reasoning), explains why APIs alone cannot be consumed directly by AI agents, and walks through how custom MCP servers for Gmail Calendar Slack and Notion let Claude Cowork and OpenWork speak to all of them equally. You will leave understanding the MCP architecture, the MCP host and MCP client relationship, and why MCP is the foundation of enterprise agentic AI.

If you want to learn:

- How to choose the best LLM for your AI agent work in 2026

- How the Arena.ai leaderboard ranks LLMs through real user voting

- Why Claude Opus 4.7 and Opus 4.6 lead the LLM rankings right now

- What the trade-off is between proprietary and open-source LLMs

- How models from Anthropic Google OpenAI xAI and DeepSeek compare on the leaderboard

- When to optimize for cost versus model quality on LLM-powered AI agents

Then this lecture is for you!

Learn how to pick the best LLM for your AI agent work using the Arena.ai leaderboard, where thousands of users vote on real model responses and produce an Elo-style ranking. The lecture walks through the leaderboard, shows why Claude Opus 4.7 and Opus 4.6 sit at the top, contrasts proprietary models from Anthropic Google OpenAI Grok with MIT-licensed open-source models like GLM Kimi and DeepSeek, and gives a practical rule of thumb: start with the best commercial model to learn fast, then optimize for cost once your workflow is solid. Apply the same logic across Claude Cowork and OpenWork to match the right LLM to each AI agent use case.

If you want to learn:

- The core prompt engineering best practices from Anthropic and OpenAI

- How zero-shot one-shot and few-shot prompting techniques differ in practice

- Why structuring prompts with Markdown and XML tags improves AI agent output

- How giving the AI a role with a system-style headline boosts response quality

- Why context engineering is overtaking pure prompt engineering for LLM agents

- Where to find the official Anthropic and OpenAI prompt engineering guides

Then this lecture is for you!

Get the high-leverage prompt engineering best practices that matter most for driving Claude Cowork and OpenWork agents reliably. The lecture distills guidance from the Anthropic Cookbook and the OpenAI prompt engineering guide on zero-shot one-shot and few-shot prompting, message formatting with Markdown and XML tags like user query and assistant response, and giving the LLM a role to anchor its reasoning. It also reframes the field through context engineering — designing the right examples and data alongside the prompt — and points to the official Anthropic and OpenAI documentation for deeper study after the course.

If you want to learn:

- How to break past the limits of native connectors in Claude Cowork and OpenWork

- How AI agents can use your browser to act on any website or app

- What end-to-end lead outreach automation looks like with an AI agent

- Why browser-based AI agents work even without a native connector

- How a single command can drive an agent across Google Sheets web Gmail Notion and Slack

Then this lecture is for you!

Make your AI agents nearly limitless by adding browser control to Claude Cowork and OpenWork. The lecture frames the section's challenge: native connectors only cover what they expose, but a lot of work lives on web pages or in tools without an MCP server. The fix is letting your AI agent drive a browser end to end. You will preview the section's flagship project: from a Google Sheet of leads, the agent opens each company website, researches it, drafts personalized outreach emails in Gmail, writes a Notion report, and posts a summary to a Slack channel — all from one autonomous AI agent command.

If you want to learn:

- How to chain multi-step AI agent actions across Notion and Slack

- How to turn a meeting prep brief into a shared Notion page automatically

- How an AI agent picks the right Slack channel for a given project context

- The difference between Claude Cowork posting as you and OpenWork posting as a bot

- How AI agents move from single actions to multi-step workflow automation

Then this lecture is for you!

Move your AI agents beyond single actions and into multi-step workflow automation by chaining Notion page creation with a Slack channel post in one command. The lecture takes the Sarah Chen meeting brief from the previous section and tells both Claude Cowork and OpenWork to publish it as a Notion page and share it with the Project Phoenix Slack channel — all in a single instruction. You will see how Claude Cowork posts on your Slack account while the OpenWork bot posts under its own identity, and learn why multi-environment actions across Notion Slack and Gmail are the natural next step in agentic AI workflows.

If you want to learn:

- How to install and enable the Claude for Chrome browser extension

- How to allow browser control and computer use inside Claude Cowork

- How AI agents use browser screenshots to navigate any web page

- How Claude in Chrome opens a new tab to run an AI agent task autonomously

- How browser control compares between Claude Cowork and OpenWork by default

Then this lecture is for you!

Set up Claude for Chrome from claude.com/claudeforchrome to give your Claude Cowork AI agent full browser control over Google Chrome. The lecture installs the Chrome extension, enables it in Cowork Customize Connectors as Claude in Chrome, flips the Browser Actions and Computer Use toggles in settings, and proves it works by asking Claude to open NASA.gov, take screenshots, and summarize the latest news. You will also see how the same browser automation works out of the box in OpenWork through its built-in Control Chrome plugin — opening the door to AI agent workflows on any web page that lacks an MCP or native connector.

If you want to learn:

- How to use OpenWork's built-in Control Chrome plugin for browser automation

- How to give an open-source AI agent full browser access without extra setup

- How OpenWork takes screenshots and reads web pages with no MCP server needed

- Why browser control unlocks AI agent tasks on sites without an MCP connector

- How OpenWork browser automation compares to Claude for Chrome in Claude Cowork

Then this lecture is for you!

Drive Google Chrome from OpenWork using the built-in Control Chrome plugin that ships enabled by default — no extension install, no API key, no separate MCP server. The lecture opens OpenWork Settings, confirms Control Chrome is connected, and runs the same NASA.gov news task to verify the AI agent can launch a Chrome tab, take screenshots, and summarize page content end to end. You will see how OpenWork browser automation matches Claude for Chrome in Claude Cowork on a cosmetic detail or two, and why this unlocks workflow automation against any web page or web app without a dedicated AI agent connector.

If you want to learn:

- How to build an end-to-end lead outreach AI agent with Claude Cowork

- How AI agents read Google Sheets through the browser without a native connector

- How to generate personalized outreach emails grounded in each company's website

- How an AI agent posts a Notion report and Slack channel update in one task

- How browser control plus Gmail Notion Slack makes one prompt do hours of work

- How AI agents fail gracefully when target websites do not exist

Then this lecture is for you!

Run the pinnacle Claude Cowork project of the course: an end-to-end lead outreach AI agent that reads a Google Sheet lead list through the browser, researches each company website, drafts personalized Gmail emails introducing a Document Processing Agent product, writes a full Notion report, and posts a summary to the sales Slack channel — all from one autonomous prompt. The lecture demonstrates browser control reading the spreadsheet, agent planning across web Gmail Notion Slack, graceful failure on fictional company sites, and the ROI of replacing hours of sales development work with a single AI agent run.

If you want to learn:

- How to connect Anthropic Claude Opus to OpenWork via the Anthropic API

- Why visual AI agent tasks need a multimodal LLM beyond BigPickle

- How to create an Anthropic API key and add credits at platform.cloud.com

- How to register Anthropic as a new model provider inside OpenWork

- How much it typically costs to run AI agent visual tasks via Claude Opus

Then this lecture is for you!

Connect Anthropic Claude Opus to OpenWork via the Anthropic API so your open-source AI agent can take screenshots, read web pages, and run visual browser tasks the free BigPickle model cannot handle. The lecture walks through platform.cloud.com to top up API credits, create a new Anthropic API key, then opens OpenWork Settings, adds Anthropic as a new model provider, pastes the key, and saves. With Claude Opus connected, your OpenWork AI agent now has the same visual reasoning power as Claude Cowork for end-to-end outreach, browser control, and any multimodal automation that needs more than a text-only LLM.

If you want to learn:

- How to run an end-to-end lead outreach AI agent inside OpenWork

- How to power an OpenWork task with Claude Opus 4.7 via the Anthropic API

- How an OpenWork AI agent handles graceful failure on fictional company sites

- How OpenWork drafts Gmail outreach emails, Notion reports, and Slack messages

- Why pairing OpenWork with Claude Opus closes the gap with paid Claude Cowork

Then this lecture is for you!

Run the same end-to-end lead outreach workflow inside OpenWork using Claude Opus 4.7 from the Anthropic API for visual browser control. The lecture switches the OpenWork model to Opus, replays the multi-company outreach task, watches the AI agent attempt to open each lead's website and gracefully prompt for clarification when fictional companies fail to resolve, then continues by drafting personalized Gmail outreach emails, posting a summary to the sales Slack channel, and writing a Notion report. The takeaway: with a small API spend, OpenWork matches the autonomous performance of Claude Cowork for real outbound sales automation.

If you want to learn:

- Why visual LLMs are required when AI agents navigate the web

- How browser use removes the memory and tools ceiling for AI agents

- Which AI agent architecture layer becomes the bottleneck after browser control

- Why reasoning is the next frontier for AI agent quality

- How Claude Opus compares to BigPickle on multimodal AI agent workflows

Then this lecture is for you!

Recap the section through the four-pillar AI agent architecture: a visual LLM like Claude Opus replaces text-only BigPickle so the agent can take screenshots and navigate the web, memory expands beyond Slack Notion Gmail Calendar to the entire internet through browser use, and tools include clicking scrolling and reading any web page. The lecture shows that with these three pillars unblocked, the next bottleneck for AI agent quality is reasoning — the still-manual instructions you copy-paste into every task. That sets up the next section's deep dive into reusable AI skills and prompt-free agentic automation.

If you want to learn:

- How LLMs process images alongside text using multimodal tokenization

- How RGB pixel values get turned into tokens for a large language model

- Why modern multimodal LLMs can read web pages with images and buttons

- How vision transformers and cross-attention join text and image embeddings

- Why multimodal LLMs are the foundation of AI agents that navigate the web

Then this lecture is for you!

Understand how large language models process images through multimodal tokenization — the same principle that makes text tokenization work, extended to pixels and UI elements. The lecture explains how RGB pixel values become numeric tokens, how mixed sequences of words, image patches, buttons, and scroll elements are fed into a single transformer architecture, and how this enables modern multimodal LLMs to read web pages and predict the next user action like a button click. You will leave understanding why visual LLMs are essential for AI agents that interact with browsers and any visual interface.

If you want to learn:

- What an AI skill is and why Anthropic introduced the concept

- How a Claude skill differs from an MCP server connector

- What the structure of a SKILL.md file looks like

- Why packaging reasoning as reusable AI skills beats copy-pasting prompts

- How skills fit into the LLM memory tools and reasoning architecture

- Where to find The Complete Guide to Building Skills for Claude

Then this lecture is for you!

Get the foundations of AI skills — Anthropic's structured format for packaging reasoning into reusable folders you and your AI agents can call by name. The lecture frames the problem with copy-pasting prompts, explains why a skill folder with a SKILL.md markdown file, optional Python scripts, references, and assets is a more durable way to express reasoning, and draws the kitchen-versus-recipe analogy: MCP servers give your AI agent ingredients and tools, while skills are the actual recipes. It also points to The Complete Guide to Building Skills for Claude and sets up the next videos on using the built-in Skill Creator in Claude Cowork and OpenWork.

If you want to learn:

- How to create your first AI skill in Claude Cowork with the Skill Creator

- How to use /skillcreator to generate a reusable Claude skill from a short prompt

- How Claude asks clarifying questions to make your skill more specific

- How the Skill Creator tests your new AI skill on real Gmail Slack Calendar Notion data

- How to save and run a custom Claude skill like a morning briefing on demand

Then this lecture is for you!

Create your first reusable AI skill inside Claude Cowork using the built-in /skillcreator command. The lecture starts from a one-line goal — generate a morning briefing skill — and shows how Claude interrogates you on what counts as overnight, urgent, VIP sender, and Notion task assignment to produce a far more specific SKILL.md than you could write by hand. You will watch the Skill Creator test the skill against live Gmail Slack Calendar and Notion data, iterate on the output, save the Morning Briefing skill under Customize Skills, and re-run it later with a single slash command. The result is a reusable Claude skill that replaces copy-pasted prompts forever.

If you want to learn:

- How to create AI skills in OpenWork using its built-in Skill Creator

- How to invoke /skillcreator inside an OpenWork session

- Why an Anthropic model is the best choice when creating Claude-style skills in OpenWork

- How a Morning Briefing skill ends up structured as a folder with scripts and instructions

- How OpenWork skill creation differs from the more interactive Claude Cowork flow

Then this lecture is for you!

Create reusable AI skills inside OpenWork using its built-in Skill Creator, the open-source counterpart to the Claude Cowork experience. The lecture opens a new OpenWork session, switches to an Anthropic model for best skill creation quality, runs /skillcreator with the same initial Morning Briefing prompt, and walks through the resulting folder containing the SKILL.md instruction file plus supporting scripts. You will also see how OpenWork skips Cloud Cowork's interactive clarifying questions and produces a one-shot skill — which requires more manual review but still ships a working /morningbriefing slash command you can re-run inside OpenWork.

If you want to learn:

- How to schedule AI skills to run automatically in Claude Cowork and OpenWork

- How to create a Claude Cowork scheduled task with a single chat command

- How OpenWork uses macOS cron jobs to schedule recurring AI agent tasks

- How to deliver a morning briefing every day at 8 a.m. without typing anything

- How to test scheduled AI agent runs and view them in the Scheduled tab

- Why scheduled AI skills eliminate yourself from every repetitive workflow

Then this lecture is for you!

Schedule your AI skills to run automatically in Claude Cowork and OpenWork so the morning briefing — or any other reusable skill — fires on a daily, weekly, hourly, or monthly cadence without your input. The lecture shows how Claude Cowork's chat recognizes a scheduling intent, loads the Create Scheduled Task tool, and adds your skill to the Scheduled tab with a Run Now option for testing. It then mirrors the same flow in OpenWork, which uses a macOS cron job to schedule the recurring task on your computer. The takeaway: connect data, build a skill, schedule it, and remove yourself from repetitive AI agent workflow execution entirely.

If you want to learn:

- How to enrich your Morning Briefing AI skill with meeting attendee research

- How to add LinkedIn and Google research to an AI agent skill workflow

- How to iterate on existing Claude or OpenWork skills directly from chat

- What other recurring workflows you can automate with AI skills next

- Why custom AI skills are the highest-leverage productivity move after this course

Then this lecture is for you!

Take the Morning Briefing AI skill one step further by adding automatic research on every meeting attendee — names, LinkedIn profiles, companies, and signals like education or certifications. The lecture walks through the manual workflow of looking up Igor Sidorenko at Neurons Lab as the motivating example, then frames it as a do-it-yourself exercise: extend your Morning Briefing skill from the chat in either Claude Cowork or OpenWork so future briefings show who you are about to meet and why they matter. It closes with prompts for other high-leverage AI agent skills across sales, marketing, project management, and team Slack summaries.

If you want to learn:

- What Claude Cowork plugins are and how they differ from individual skills

- How to package your AI agent skills as plugins for your team or the world

- Why plugins are the right unit for sharing AI agent expertise at scale

- How Claude Cowork and OpenWork support plugin packaging

- What's coming next in the section on plugins for AI agents

Then this lecture is for you!

Move from building AI agents for yourself to sharing them with the world by packaging your skills as plugins. The lecture frames the section: after data connectors, browser control, custom skills, and scheduled tasks, the natural next step is bundling that reasoning into Claude Cowork plugins or OpenWork shareable links so colleagues and the wider community can install your work. You will preview how plugins package multiple skills plus connectors into a single distributable unit and see the path to publishing your custom AI agent expertise across Claude Cowork plugins and OpenWork plugin marketplaces in the videos that follow.

If you want to learn:

- How to find and install plugins in Claude Cowork

- Where to browse official Anthropic and partner plugins inside Cowork

- Why the Anthropic Sales plugin is a great place to start for AI agent workflows

- What skills and connectors live inside a Claude Cowork plugin

- How to use the call prep slash command from the installed Sales plugin

- Why third-party plugin marketplaces require extra security caution

Then this lecture is for you!

Install your first Claude Cowork plugin and use it to power real AI agent workflows. The lecture walks through Customize and Browse Plugins to find the official Anthropic Sales plugin, installs it in one click, and explores the bundled skills — call prep, call summary, draft outreach, forecast, pipeline review — plus the wider set of connectors it ships with. You will then run the call prep skill against a meeting on your calendar and watch Claude Cowork stitch Gmail Calendar and Notion data into a meeting brief, while learning why third-party plugin marketplaces like claude marketplaces.com deserve more security scrutiny than the official Anthropic catalog.

If you want to learn:

- How to use the Anthropic Sales plugin's call prep skill in Claude Cowork

- What a high-quality AI meeting brief from a plugin skill looks like

- How a Claude Cowork plugin combines Gmail Calendar and Notion data for sales prep

- What discovery questions and objection handling AI agents can surface for sales calls

- Why trusting Anthropic-developed plugins extends your AI agent skills safely

Then this lecture is for you!

See the Anthropic Sales plugin's call prep skill in action inside Claude Cowork as it produces a smarter meeting brief than the custom skills you built earlier in the course. The lecture opens the generated brief for a meeting with Igor Sidorenko at Neurons Lab, walks through the company overview, LinkedIn-backed background, decision-maker context, reply history, suggested agenda, discovery questions, and pre-written objection handling — none of which you wired up yourself. The takeaway: trusted Anthropic plugins extend your AI agent's intelligence beyond your own custom skills while still using your existing Gmail Calendar Notion data.

If you want to learn:

- How to install OpenWork plugins from the Awesome OpenCode GitHub list

- Why OpenWork plugins are more technical than Claude Cowork plugins today

- How to install a Context Analysis plugin to track AI agent token usage

- How to clone an OpenCode plugin repo and copy it into your OpenWork project

- How to use the /context slash command after installing an OpenWork plugin

- Why monitoring token usage matters when running paid Anthropic models in OpenWork

Then this lecture is for you!

Install third-party OpenWork plugins from the GitHub Awesome OpenCode list to extend your open-source AI agent beyond the built-in capabilities. The lecture explains why OpenCode plugins are mostly aimed at software engineers, then walks through cloning a Context Analysis plugin repo, copying its opencode folder into your OpenWork project folder, reloading OpenWork, and running the new /context slash command. You will see exactly how your tokens are split across system, user message, Slack, Notion, and other tools — essential telemetry when you are paying for Claude Opus or any other API-backed LLM inside OpenWork.

If you want to learn:

- How to create your own Claude Cowork plugin from existing skills

- How Claude's plugin builder asks who the plugin is for and packages it accordingly

- What a Claude Cowork plugin folder contains: skills, schedules, connectors, README

- How to add a scheduled task inside a plugin manifest for recurring AI agent runs

- How to share a Claude Cowork plugin with your team or publish it for the world

Then this lecture is for you!

Create your own Claude Cowork plugin from an existing skill — for example, a Morning Briefing — and learn how to share it with your team or the wider world. The lecture opens Customize Plugins, clicks Create plugin, and asks Claude to build a public Morning Briefing plugin using the skill you already wrote. Claude's plugin builder interrogates audience, plugin name, scheduled task setup, and manifest authorship, then assembles a folder containing the skill, a setup schedule, the connectors list (email, chat, calendar, notes, tasks), and a README. You will then see how to open the working folder, archive it, and distribute the plugin to teammates or the public.

If you want to learn:

- How to share OpenWork AI skills with teammates without packaging a plugin

- Where local OpenWork skills live on your computer in .opencode/skills

- How to use OpenWork's Share Skill feature to publish a public link

- Why sharing skills in OpenWork is easier than authoring full plugins

- How a single SKILL.md file plus optional scripts is enough to share an AI agent skill

Then this lecture is for you!

Share your custom OpenWork AI skills with teammates or the world without ever packaging them as full plugins. The lecture opens the local Skills folder under .open code/skills, shows that a skill is just a SKILL.md text file with optional scripts, and walks through two distribution paths: copy the skill folder directly into a teammate's setup, or use OpenWork's built-in Share Skill action to create either an organization-only link or a public sharable URL anyone can install. The result is a low-friction way to spread the Morning Briefing skill and the rest of your AI agent expertise across teams while plugin authoring stays out of scope.

If you want to learn:

- What context engineering is and how it differs from prompt engineering

- Why context engineering matters more than prompt engineering for AI agents

- How to decide which documents tools and connectors belong in your context window

- Why too much context causes hallucinations and burns tokens

- How skills and plugins encode context engineering decisions for you

- Where to read Anthropic's official article on context engineering

Then this lecture is for you!

Move past pure prompt engineering and into context engineering — the higher-leverage discipline of deciding exactly which documents tools connectors memories and instructions land in an AI agent's context window for each task. The lecture summarizes Anthropic's article on context engineering, explains why production AI systems care about the right context rather than the longest prompt, links context engineering directly to skills and plugins as the practical place where these decisions live, and warns about token spend and hallucinations when you over-pack the context window. Apply it to design cost-efficient agentic AI workflows in Claude Cowork and OpenWork.

If you want to learn:

- What the four bonus Claude applications are: code design dashboards and visualizations

- How the Claude ecosystem extends AI agent productivity beyond Claude Cowork

- Why visual and software-product Claude tools compound your existing skills and plugins

- What's coming next in this bonus section of the course

Then this lecture is for you!

Get oriented to the bonus section that compounds everything you have learned so far in the course. Now that you have mastered AI agent fundamentals — data connectors, browser control, reusable AI skills, scheduled tasks, and plugins — this short intro previews four Claude ecosystem applications that extend your productivity even further: writing code, generating design, creating live dashboards, and producing visualizations. Each of the next four videos walks through one of these tools so you can see how the broader Claude ecosystem layers on top of Claude Cowork.

If you want to learn:

- How to create static visualizations inside Claude Cowork chat

- How to ask Claude for tree-based diagrams of negotiation tactics

- How AI agents combine Gmail Slack and web research into visual playbooks

- How to save Claude-generated visualizations as files for reuse

- Why visualizations help structure information from AI agent outputs

Then this lecture is for you!

Use Claude Cowork chat to generate static visualizations that turn AI agent research into clean visual playbooks. The lecture builds on the call prep skill, asking Claude for a tree-based visualization of negotiation tactics for a specific contact, while letting the AI agent pull live context from Gmail Slack Calendar and the web. The result is a structured negotiation diagram covering rapport, framing, exchange, and closing — copyable to clipboard or saved as a file. The takeaway: visual thinkers can lean on Claude Cowork to compress AI agent research into instantly scannable visualizations.

If you want to learn:

- How to build live dashboards in Claude Cowork using dynamic Live Artifacts

- How Live Artifacts differ from static visualizations in Claude

- How to wire Gmail Calendar Slack and Notion data into a refreshable HTML dashboard

- How to set up a prioritized daily command center grouped by source and urgency

- How to refresh a Claude Cowork dashboard with new emails tasks and Slack messages

Then this lecture is for you!

Build dynamic, refreshable dashboards inside Claude Cowork using Live Artifacts — HTML pages that re-pull data from your connectors every time you open them. The lecture opens Live Artifacts, picks the daily command center template, connects Gmail Calendar Slack and Notion, and answers Claude's design questions on grouping, urgency, and time window. With one Create click, Claude Cowork produces an HTML dashboard that you can reload to pull fresh emails, tasks, calendar events, and Slack messages. The result is a custom AI-powered command center sitting in the Live Artifacts tab that complements the static visualizations covered earlier.

If you want to learn:

- How to build a high-fidelity HTML mockup with Claude Code from your connected data

- How Claude Code differs from Claude Cowork and the Claude chatbot

- How an AI coding agent uses Notion Slack Gmail and Calendar context to generate a UI

- How product managers and non-developers can ship clickable demos with Claude Code

- How Claude Code uses your existing skills plugins and connectors out of the box

- How to test a Claude Code mockup on both desktop and mobile views

Then this lecture is for you!

Use Claude Code — the third app in the Claude desktop bundle — to turn everything in your Notion Slack Gmail and Calendar about a real project into a high-fidelity HTML mockup you can show to clients or investors. The lecture frames Claude Code as Claude Cowork's coding sibling that shares all your skills, plugins, and connectors, runs a Project Phoenix mockup task, watches the AI agent gather context across data sources and write a single-file HTML app with navigation and dashboards, and then opens it locally with desktop and mobile views. The result: product managers and non-developers can ship clickable prototypes in minutes.

If you want to learn:

- How to create beautiful slide decks with Claude AI Design from your connected data

- How claude.ai/design uses Notion Slack Gmail and Calendar to draft presentations

- How to give Claude AI Design a design system, logos, fonts, and Figma assets

- How Claude AI Design compares to Claude Code for generating client-ready deliverables

- Why a status report deck can be auto-generated for investors in minutes

Then this lecture is for you!

Create polished, investor-ready slide decks with Claude AI Design — the web feature at claude.ai/design that turns your connected data into beautiful presentations. The lecture walks through creating a Project Phoenix Status Report deck, importing optional design system assets like logos fonts colors and Figma files, and asking Claude to draft a full investor presentation using Notion Slack Gmail and Calendar context. You will see Claude AI Design pull headlines about deal signing, document type distributions, AI agent architecture, and project expansion plans — then output styled slides in minutes that complement Claude Code and Claude Cowork in the broader Claude ecosystem.

If you want to learn:

- What products and plugins live inside the Anthropic ecosystem in 2026

- How Claude for Excel embeds Claude directly into Microsoft 365 spreadsheets

- Why Claude for PowerPoint matters for client-facing presentation workflows

- How Claude for Slack brings AI assistants into team workspaces

- Where Claude Security Claude Code and Claude Cowork sit alongside add-in plugins

- How to identify which Anthropic plugin best matches your daily working environment

Then this lecture is for you!

Get a final tour of the wider Anthropic ecosystem beyond the Claude Cowork desktop app: Claude Code for software work, Claude Security for vulnerability scanning, and specialized plugins for Claude for Slack, Claude for Excel, Claude for PowerPoint, and Claude for Word that live inside Microsoft 365 and Slack rather than in Cowork. The lecture explains why a finance or accounting professional benefits from Claude living inside Excel directly, why presentation-heavy roles benefit from Claude for PowerPoint, and how to explore Anthropic's official Meet Claude product line to find the right Claude integration for your daily working environment.