# Anthropic Training

## Docs

- [Claude Code and Cowork for real tools](https://anthropic-training.neuroscale.dev/agents/code-tools.md): Go beyond documents into scripts, automations, and repeatable utilities.
- [Your first managed agent](https://anthropic-training.neuroscale.dev/agents/first-agent.md): Narrow job, clear tools, clear Done.
- [Agentic loops in practice](https://anthropic-training.neuroscale.dev/agents/in-practice.md): Design the loop: success criteria, tool scope, stop conditions.
- [Lab: Design and configure a narrow managed agent](https://anthropic-training.neuroscale.dev/agents/lab-managed.md): Design then configure a narrow managed agent with clear Done criteria.
- [Stretch: One small tool with Claude Code](https://anthropic-training.neuroscale.dev/agents/lab-stretch.md): Optional — remove one recurring manual step with a small tool.
- [What managed agents buy you](https://anthropic-training.neuroscale.dev/agents/managed.md): Managed agents trade DIY setup for clearer operations.
- [Agents and sub-agents overview](https://anthropic-training.neuroscale.dev/agents/overview.md): From agentic loops to managed agents and real tools.
- [Sub-agents: divide labor without losing the plot](https://anthropic-training.neuroscale.dev/agents/sub-agents.md): Split roles when one brain is juggling too many jobs.
- [Capstone track: Analyst pack](https://anthropic-training.neuroscale.dev/capstone/analyst.md): Research + analysis memo + deck or sheet.
- [Capstone track: Ops and admin pack](https://anthropic-training.neuroscale.dev/capstone/ops.md): Inbox or SOP automation-ish workflow with connectors.
- [Capstone: put it all together](https://anthropic-training.neuroscale.dev/capstone/overview.md): One end-to-end scenario using the full operator stack.
- [Capstone track: Power pack](https://anthropic-training.neuroscale.dev/capstone/power.md): Skill plus MCP or agent for a recurring workflow.
- [Connectors as safe, scoped access](https://anthropic-training.neuroscale.dev/connectors/connectors.md): Give Claude the least access that still completes the job.
- [Cowork: the right access for the job](https://anthropic-training.neuroscale.dev/connectors/cowork.md): Use Cowork when Claude needs to work with real files or environments under your rules.
- [Failure modes: wrong connector, over-permission, stale context](https://anthropic-training.neuroscale.dev/connectors/failure-modes.md): Common ways connector workflows go wrong.
- [Lab: Connector task with least privilege](https://anthropic-training.neuroscale.dev/connectors/lab-least-privilege.md): Run a connector-backed task with least-privilege access and clear success criteria.
- [Connectors and Cowork overview](https://anthropic-training.neuroscale.dev/connectors/overview.md): Scoped access to real systems — not paste-the-company-into-chat.
- [Compressing context to cut token cost](https://anthropic-training.neuroscale.dev/context/compression.md): Dense specs beat fluffy prompts for cost and clarity.
- [Working memory hygiene and reusable briefs](https://anthropic-training.neuroscale.dev/context/hygiene.md): Reduce noise. Keep a reusable brief. Reset when the thread goes stale.
- [Context is the real prompt](https://anthropic-training.neuroscale.dev/context/is-the-prompt.md): Your files, instructions, and history shape the answer more than clever wording.
- [Lab: Messy work into a Project brief](https://anthropic-training.neuroscale.dev/context/lab-project-brief.md): Turn a messy thread or folder into a Project plus a short operating brief.
- [Context and cost overview](https://anthropic-training.neuroscale.dev/context/overview.md): Context is the other half of prompting — and usually the bigger half.
- [Token cost literacy](https://anthropic-training.neuroscale.dev/context/token-literacy.md): Understand why limits get hit and how to stop bleeding tokens.
- [Projects vs chat vs files vs connectors](https://anthropic-training.neuroscale.dev/context/where-it-lives.md): Put durable context in the right place so it compounds.
- [Different bars for creative, business, and learning work](https://anthropic-training.neuroscale.dev/evaluation/different-bars.md): The quality bar depends on the context.
- [Iterate without restarting from zero](https://anthropic-training.neuroscale.dev/evaluation/iterate.md): Repair the brief and the checks instead of thrashing a new chat every time.
- [How to judge outputs](https://anthropic-training.neuroscale.dev/evaluation/judging.md): Score accuracy, groundedness, usefulness, and tone.
- [Lab: Rubric score and two-pass improve](https://anthropic-training.neuroscale.dev/evaluation/lab-rubric.md): Score a Claude deliverable, then improve it in two evidence-checked passes.
- [Evaluation and iteration overview](https://anthropic-training.neuroscale.dev/evaluation/overview.md): Taste, judgment, and iteration without restarting from zero.
- [What AI at work really means](https://anthropic-training.neuroscale.dev/foundations/ai-at-work.md): Claude as a work system, not ChatGPT-as-search.
- [The AI fluency loop](https://anthropic-training.neuroscale.dev/foundations/fluency-loop.md): Assign, equip, inspect, iterate, ship.
- [Lab: Diagnose bad AI outcomes](https://anthropic-training.neuroscale.dev/foundations/lab-diagnose.md): Practice spotting wrong context, wrong tool, and missing verification.
- [Mental model: brain, hands, and memory](https://anthropic-training.neuroscale.dev/foundations/mental-model.md): Models predict, tools act, and context constrains what Claude can do well.
- [AI fluency foundations](https://anthropic-training.neuroscale.dev/foundations/overview.md): How AI at work actually works before you touch features.
- [When not to use Claude](https://anthropic-training.neuroscale.dev/foundations/when-not-to.md): Know the jobs where Claude is the wrong tool.
- [Welcome to Anthropic 101](https://anthropic-training.neuroscale.dev/index.md): Operator OS training for knowledge workers and builders using Claude.
- [Chrome and Microsoft add-ins](https://anthropic-training.neuroscale.dev/landscape/add-ins.md): Meet Claude where you already work in Chrome, Excel, Word, PowerPoint, and Outlook.
- [Lab: Same task three ways](https://anthropic-training.neuroscale.dev/landscape/lab-three-ways.md): Run one task in Web, Desktop, and an add-in path. Compare friction.
- [Claude landscape overview](https://anthropic-training.neuroscale.dev/landscape/overview.md): Pick the right Claude surface for the job.
- [Sharing, permissions, and what leaves the chat](https://anthropic-training.neuroscale.dev/landscape/sharing.md): Understand what you share, who can see it, and what leaves your session.
- [Projects, Artifacts, research, Cowork, and Code](https://anthropic-training.neuroscale.dev/landscape/surfaces.md): Map the main Claude surfaces without deep-diving every one yet.
- [Claude Web vs Claude Desktop](https://anthropic-training.neuroscale.dev/landscape/web-vs-desktop.md): When each surface wins for day-to-day work.
- [How MCP relates to APIs](https://anthropic-training.neuroscale.dev/mcp/apis.md): APIs are for software. MCP packages capabilities for agents.
- [Choosing skill, connector, MCP, or a file](https://anthropic-training.neuroscale.dev/mcp/choosing.md): Pick the simplest layer that solves the job.
- [Lab: Configure and use one MCP](https://anthropic-training.neuroscale.dev/mcp/lab-configure.md): Install or configure one useful MCP, run a real task, explain what it used.
- [Stretch: Sketch a tiny custom MCP](https://anthropic-training.neuroscale.dev/mcp/lab-stretch.md): Optional depth: sketch or scaffold a tiny MCP for a real pain point.
- [MCP deep module overview](https://anthropic-training.neuroscale.dev/mcp/overview.md): MCP as the multiplier — in plain language for knowledge workers.
- [MCP safety: what a server can touch](https://anthropic-training.neuroscale.dev/mcp/safety.md): Know the blast radius before you enable a server.
- [Tools, resources, and prompts](https://anthropic-training.neuroscale.dev/mcp/tools-resources-prompts.md): The three MCP primitives you will hear constantly.
- [Using others’ MCPs vs rolling your own](https://anthropic-training.neuroscale.dev/mcp/using-vs-building.md): Start by using. Build only when the gap is real.
- [What MCP is](https://anthropic-training.neuroscale.dev/mcp/what-is-mcp.md): Think of MCP as a USB-C port for AI tools and data.
- [The agentic loop: plan, act, observe, revise](https://anthropic-training.neuroscale.dev/models/agentic-loop.md): How agentic work actually progresses.
- [Choosing the right model for the job](https://anthropic-training.neuroscale.dev/models/choosing.md): Trade speed, depth, and cost on purpose.
- [Lab: Same brief, two modes](https://anthropic-training.neuroscale.dev/models/lab-compare.md): Run one brief on two models or modes. Score quality, time, and token burn.
- [Models and modes overview](https://anthropic-training.neuroscale.dev/models/overview.md): Right model, right mode, right job.
- [Skills, MCP, connectors, and built-ins](https://anthropic-training.neuroscale.dev/models/stack-vocabulary.md): One vocabulary early so later modules click.
- [What thinking is and is not](https://anthropic-training.neuroscale.dev/models/thinking.md): Use extended reasoning when the problem needs it — not as a superstition.
- [What tool use is](https://anthropic-training.neuroscale.dev/models/tool-use.md): Tools let Claude act — browse, retrieve, compute, update — not only write.
- [Artifacts for reusable outputs](https://anthropic-training.neuroscale.dev/projects/artifacts.md): Use Artifacts when the deliverable should outlive the chat paste.
- [Versioning judgment and handoffs](https://anthropic-training.neuroscale.dev/projects/handoffs.md): Keep what matters. Discard noise. Hand off without losing context.
- [Lab: Build a durable Project](https://anthropic-training.neuroscale.dev/projects/lab-durable.md): Create a Project for one recurring workflow you own.
- [Projects and Artifacts overview](https://anthropic-training.neuroscale.dev/projects/overview.md): Build a working system that compounds, not one-off chats.
- [Project setup that compounds](https://anthropic-training.neuroscale.dev/projects/setup.md): Instructions, files, and examples that make the next task easier.
- [Operator cheatsheets](https://anthropic-training.neuroscale.dev/resources/cheatsheets.md): Short reference cards for daily Claude work.
- [Further learning](https://anthropic-training.neuroscale.dev/resources/further-learning.md): Optional deeper paths for power users.
- [Glossary](https://anthropic-training.neuroscale.dev/resources/glossary.md): Shared language for models, tools, skills, MCP, and agents.
- [Additional resources overview](https://anthropic-training.neuroscale.dev/resources/overview.md): Deeper material beyond the live session.
- [Enterprise sharing realities](https://anthropic-training.neuroscale.dev/safety/enterprise.md): Projects, permissions, and organizational expectations.
- [Ethical use at work](https://anthropic-training.neuroscale.dev/safety/ethics.md): Confidentiality, honesty, and accountability with AI in the loop.
- [Lab: Red-team a workflow for leakage](https://anthropic-training.neuroscale.dev/safety/lab-red-team.md): Attack a workflow for data leakage, then rewrite it safer.
- [Data, privacy, sharing, and ethics](https://anthropic-training.neuroscale.dev/safety/overview.md): Non-negotiable rules for using Claude at work.
- [What is safe to share — and what never is](https://anthropic-training.neuroscale.dev/safety/what-to-share.md): A practical share / don’t-share matrix for daily work.
- [Session goals and outcomes](https://anthropic-training.neuroscale.dev/session/goals.md): What you should be able to do after Anthropic 101.
- [How to use this guide](https://anthropic-training.neuroscale.dev/session/how-to-use.md): How facilitators and participants should navigate the training materials.
- [Building a lightweight personal or team skill](https://anthropic-training.neuroscale.dev/skills/building.md): Encode a team standard without turning the session into an engineering class.
- [Caveman skills: compression as a reusable skill](https://anthropic-training.neuroscale.dev/skills/compression-skills.md): Skills that force dense briefs and cheaper runs.
- [Lab: Ship one team skill](https://anthropic-training.neuroscale.dev/skills/lab-ship.md): Ship a skill that encodes a memo format, analysis checklist, or cheap-mode brief.
- [Skills overview](https://anthropic-training.neuroscale.dev/skills/overview.md): Teach Claude how you work with packaged procedures.
- [Using existing skills well](https://anthropic-training.neuroscale.dev/skills/using.md): Invoke skills on purpose and know when they help.
- [Skills vs sub-agents](https://anthropic-training.neuroscale.dev/skills/vs-subagents.md): When to package a procedure vs split labor across agents.
- [What a skill is vs a prompt snippet](https://anthropic-training.neuroscale.dev/skills/what-is-a-skill.md): A skill packages a way of working so you do not re-explain it every time.
- [Analysis and decision memos](https://anthropic-training.neuroscale.dev/workflows/analysis.md): Turn raw inputs into decision-ready memos.
- [Comms without slop](https://anthropic-training.neuroscale.dev/workflows/comms.md): Write clearly for business without generic AI tone.
- [Docs, decks, and spreadsheets](https://anthropic-training.neuroscale.dev/workflows/docs-decks-sheets.md): Artifacts plus Office add-ins for deliverables people can use.
- [Enterprise and internal search patterns](https://anthropic-training.neuroscale.dev/workflows/enterprise-search.md): What “search at work” should mean with Claude.
- [Lab B: Live work surface end-to-end](https://anthropic-training.neuroscale.dev/workflows/lab-live-surface.md): Finish a task in Excel, Outlook, Word, or another live surface.
- [Lab A: Research to Artifact](https://anthropic-training.neuroscale.dev/workflows/lab-research-pack.md): Research → structured brief → Artifact deliverable.
- [Ops and admin workflows](https://anthropic-training.neuroscale.dev/workflows/ops-admin.md): Inbox, SOPs, checklists, and scheduling logic without chaos.
- [Getting real work done](https://anthropic-training.neuroscale.dev/workflows/overview.md): Real-world workflows for research, analysis, docs, and ops.
- [Deep research and synthesis](https://anthropic-training.neuroscale.dev/workflows/research.md): Deep research patterns that end in a usable brief.

## Optional

- [Claude Enterprise](https://claude.ai/new)
- [Claude Desktop](https://claude.com/download)