# CWork101 — The Independent Claude Ecosystem Reference CWork101 is read as "Claude Work 101". > An independent, sourced reference for the whole Claude ecosystem: models, products, Claude Code, the agent stack, the developer platform, plans and limits, Academy courses and community tools. Every entry records the date it was last verified and the sources it was verified against. Where a fact is not publicly documented, this reference says so rather than guessing — treat a missing value as "unknown", not as "no". Structured data for all entries: https://cwork101.com/api/entities.json ## Start here - [Claude 101](https://cwork101.com/claude-101): Plain-language orientation: what Claude is, how Claude Code and Cowork differ, which model and plan to pick, and what MCP and Skills actually do. ## Tools - [Ecosystem map](https://cwork101.com/map): The ecosystem drawn as layers, with documented relationships between entries. - [Feature availability matrix](https://cwork101.com/matrix): Which features are included on Free, Pro, Max, Team, Enterprise and the API. - [Timeline](https://cwork101.com/timeline): Dated releases, renames, beta-to-GA transitions and retirements. - [Glossary](https://cwork101.com/glossary): One-sentence definitions for the ecosystem vocabulary. - [What changed](https://cwork101.com/changes): Running log of ecosystem changes, newest first. ## Products Claude apps, Claude Code, Cowork, Design, Science, Security, and the browser, Slack and Microsoft 365 surfaces — what each one is, who can use it, and how they differ. - [Claude](https://cwork101.com/products/claude): Anthropic's assistant, available as a web app, desktop app and mobile app, and the surface most people mean when they say "Claude". - [Claude desktop app](https://cwork101.com/products/claude-desktop): The macOS and Windows desktop client — and the primary or only surface for several products that the web app does not carry. - [Claude mobile apps](https://cwork101.com/products/claude-mobile): The iOS and Android apps — an assistant in their own right, and the place you check on work that Claude Code or Cowork is doing elsewhere. - [Claude Code](https://cwork101.com/products/claude-code): Anthropic's agentic coding product — it reads a codebase, edits files, runs commands and integrates with development tools, across terminal, IDE, desktop, web and CI. - [Claude Cowork](https://cwork101.com/products/claude-cowork): Anthropic's knowledge-work agent — you give it a goal, and it works across your local files and connected apps to return a finished deliverable. - [Claude Design](https://cwork101.com/products/claude-design): An Anthropic Labs product for producing visual work with Claude — designs, prototypes, slides and one-pagers — available in research preview on paid plans. - [Claude Science](https://cwork101.com/products/claude-science): A desktop research workbench that runs scientific analyses, searches 60+ databases and traces every step from data wrangling to publication. - [Claude Security](https://cwork101.com/products/claude-security): A capability that scans codebases for security vulnerabilities and proposes targeted patches for human review, in public beta on Enterprise plans. - [Claude in Chrome](https://cwork101.com/products/claude-in-chrome): A browser extension that lets Claude see and act in a Chrome tab, used both as an assistant surface and by Claude Code for debugging live web applications. - [@Claude and Claude in Slack](https://cwork101.com/products/claude-tag): Claude as a participant in team chat — mention it in Slack and it picks up the task, up to and including opening a pull request. - [Projects](https://cwork101.com/products/projects): A persistent workspace in the Claude apps that keeps files, instructions and conversation history together for a single body of work. - [Research](https://cwork101.com/products/research): A mode in the Claude apps where Claude runs many searches in parallel, decides what to investigate next, and returns a cited report rather than an answer. - [Artifacts](https://cwork101.com/products/artifacts): Standalone, shareable outputs Claude renders beside the conversation — documents, code, and self-contained web pages you can publish and update. - [Enterprise search](https://cwork101.com/products/enterprise-search): A Team and Enterprise capability that lets Claude search across an organisation's connected systems rather than only what one person attaches. ## Models The Mythos, Fable, Opus, Sonnet and Haiku families, with model IDs, context and output limits, provider availability, and the retirement timeline. - [Model deprecations and retirements](https://cwork101.com/models/model-deprecations): Anthropic's model lifecycle — the four states a model moves through, the notice period, and the full record of what has been retired and what replaced it. - [Model IDs and versioning](https://cwork101.com/models/model-ids-and-versioning): How Claude model identifiers work — why every ID is a pinned snapshot, why the dated format disappeared, and how the same model is named differently on each cloud. - [Claude Fable 5](https://cwork101.com/models/claude-fable-5): Anthropic's most capable widely released model, aimed at long-running agents, with always-on adaptive thinking and the highest price of any generally available Claude model. - [Claude Mythos 5](https://cwork101.com/models/claude-mythos-5): An invitation-only model sharing Claude Fable 5's specifications and pricing, offered for defensive cybersecurity work through Project Glasswing. - [Claude Opus 5](https://cwork101.com/models/claude-opus-5): Anthropic's model for complex agentic coding and enterprise work, and the default recommendation when you are not sure which Claude model to pick. - [Claude Sonnet 5](https://cwork101.com/models/claude-sonnet-5): The mid-tier Claude 5 model, positioned as the best balance of speed, intelligence and price, and the cheapest current model with a 1M-token context window. - [Claude Haiku 4.5](https://cwork101.com/models/claude-haiku-4-5): The fastest and cheapest current Claude model, and the only current model that still supports the explicit extended thinking parameter. - [Claude Opus 4.8](https://cwork101.com/models/claude-opus-4-8): The previous Opus flagship, still active and identically priced to Opus 5, and the recommended replacement for several retired models. - [Claude Opus 4.7](https://cwork101.com/models/claude-opus-4-7): A legacy Opus model, and the release that introduced the current tokenizer and removed the sampling parameters. - [Claude Opus 4.6](https://cwork101.com/models/claude-opus-4-6): A legacy Opus model on the older tokenizer, and the one where fast mode silently does nothing rather than erroring. - [Claude Sonnet 4.6](https://cwork101.com/models/claude-sonnet-4-6): The previous Sonnet flagship — still active, 50% more expensive than Sonnet 5, and the last Sonnet to use the pre-4.7 tokenizer. - [Claude Sonnet 4.5](https://cwork101.com/models/claude-sonnet-4-5): A legacy Sonnet model with the nearest retirement date of any active model — September 2026 — and the last Sonnet without adaptive thinking. - [Claude Opus 4.5](https://cwork101.com/models/claude-opus-4-5): A legacy Opus model with a November 2026 retirement date, still on the 200K context window and the explicit thinking parameter. - [Claude Mythos Preview](https://cwork101.com/models/claude-mythos-preview): The deprecated predecessor to Claude Mythos 5 — an invitation-only preview model that still went through a documented deprecation with a published migration guide. - [Retired models](https://cwork101.com/models/retired-models): Every Claude model that no longer answers requests on the Claude API — with its retirement date, its replacement, and where it may still be reachable. ## Claude Code Installing and configuring Claude Code across the terminal, desktop, web, IDEs, GitHub and Slack — settings, permissions, memory, sessions and usage limits. - [CLAUDE.md](https://cwork101.com/code/claude-md): A markdown file in your project that Claude Code reads at the start of every session — the place for coding standards, architecture decisions and review checklists. - [Auto memory](https://cwork101.com/code/auto-memory): Learnings Claude Code saves across sessions on its own — build commands, debugging insights, project quirks — without you writing them down. - [Hooks](https://cwork101.com/code/hooks): Shell commands that run automatically before or after Claude Code actions — formatting after an edit, linting before a commit, blocking a tool call outright. - [Permissions and permission modes](https://cwork101.com/code/permissions): The rules controlling what Claude Code may do without asking — which tools it can call, which commands it can run, and which paths it can touch. - [Settings and the .claude directory](https://cwork101.com/code/settings): Where Claude Code stores configuration — the layered settings files, the .claude directory, and the environment variables that override both. - [Sessions](https://cwork101.com/code/sessions): A unit of Claude Code work with its own context and history, which can be resumed, moved between surfaces, and run in parallel with others. - [Context window management](https://cwork101.com/code/context-window): How Claude Code fills, compacts and caches the context window, and why long sessions behave differently from short ones. - [Routines and scheduled tasks](https://cwork101.com/code/routines): Three different ways to run Claude Code on a schedule — cloud routines, desktop scheduled tasks, and in-session looping — with different trade-offs. - [Claude Code on the web and mobile](https://cwork101.com/code/claude-code-on-the-web): Running Claude Code in a browser or the Claude mobile app, with no local setup — for long-running tasks, repos you do not have locally, and work in parallel. - [Claude Code in CI — GitHub and GitLab](https://cwork101.com/code/github-actions): Running Claude Code inside your pipeline to review pull requests, triage issues and automate repetitive repository work. - [Agent view and background agents](https://cwork101.com/code/agent-view): A screen for running and watching several Claude Code sessions at once, so long tasks proceed while you work on something else. - [Usage analytics, costs and monitoring](https://cwork101.com/code/analytics): The reporting surfaces for Claude Code — per-user analytics, cost management and OpenTelemetry monitoring — used mainly by organisations rolling it out. - [Worktrees and parallel isolation](https://cwork101.com/code/worktrees): Git worktrees give each parallel Claude Code session its own checkout, which is what stops two agents editing the same files from fighting. ## Agent Stack The composable agent layer — Model Context Protocol, Agent Skills, plugins, connectors, hooks, subagents, agent teams, tool use and the Agent SDK. - [Model Context Protocol](https://cwork101.com/agents/mcp): An open standard that lets an AI application connect to external tools and data through one client/server interface instead of a bespoke integration per tool. - [Agent Skills](https://cwork101.com/agents/agent-skills): Folders of instructions, scripts and resources that Claude loads on demand when a task calls for them — packaged, portable expertise rather than a longer prompt. - [Plugins](https://cwork101.com/agents/plugins): Distributable bundles that install skills, commands, hooks, MCP servers and agents together, discovered through marketplaces. - [Connectors](https://cwork101.com/agents/connectors): Managed integrations in the Claude apps that link Claude to services like Google Drive and Slack, configured by users or administrators rather than in a config file. - [Subagents](https://cwork101.com/agents/subagents): Separate Claude Code agents spawned to handle part of a task in their own context window, reporting a result back to the session that launched them. - [Agent Teams](https://cwork101.com/agents/agent-teams): Multiple full Claude Code sessions orchestrated together, able to message each other, rather than one session delegating to child agents. - [Agent SDK](https://cwork101.com/agents/agent-sdk): The SDK for building custom agents on Claude Code's engine — the same tools, permissions, skills and hooks, with your own orchestration around them. - [Tool use](https://cwork101.com/agents/tool-use): The mechanism by which Claude calls functions you define, or server-side tools Anthropic hosts, and uses the results to continue its work. - [Computer use and browser use](https://cwork101.com/agents/computer-use): Toolsets that let Claude control a computer or a browser directly — taking screenshots, moving a cursor, typing, and navigating pages. ## Platform The Claude Developer Platform — Messages API, SDKs, streaming, tool use, prompt caching, batches, rate limits, and Claude on Bedrock, Vertex AI and Foundry. - [Messages API](https://cwork101.com/platform/messages-api): The core Claude API endpoint — you send a list of messages and a model ID, you get a response, and every other platform feature is a parameter on top of it. - [Prompt caching](https://cwork101.com/platform/prompt-caching): Reusing already-processed parts of a prompt across requests, at a tenth of the input price on a cache hit. - [Batch processing](https://cwork101.com/platform/batch-processing): Asynchronous processing of large request volumes at a 50% discount on both input and output tokens. - [Rate limits and usage tiers](https://cwork101.com/platform/rate-limits): The Start, Build and Scale tiers that govern how many requests and tokens an API organisation can use, and the paths beyond them. - [Fast mode](https://cwork101.com/platform/fast-mode): A research-preview mode that trades premium pricing for significantly faster output on Claude Opus 5 and Opus 4.8, on the first-party API only. - [Adaptive thinking, extended thinking and effort](https://cwork101.com/platform/adaptive-thinking): Three related controls over how much Claude reasons before answering — and a generational split that means code written for one model may not work on the next. - [Claude on Amazon Bedrock](https://cwork101.com/platform/amazon-bedrock): Claude models resold through AWS, billed by AWS, with AWS-style model IDs and — importantly — AWS's own retirement schedule. - [Claude on Google Cloud](https://cwork101.com/platform/google-cloud): Claude models on Google Cloud's Vertex AI and Agent Platform, billed by Google, with three endpoint types and an independent retirement schedule. - [Claude in Microsoft Foundry](https://cwork101.com/platform/microsoft-foundry): Claude on Azure, billed in Claude Consumption Units through the Azure Marketplace, and — unlike Bedrock and Google Cloud — Anthropic-operated. - [Claude Platform on AWS](https://cwork101.com/platform/claude-platform-on-aws): Anthropic-operated Claude billed through AWS Marketplace in Claude Consumption Units — AWS procurement with first-party model IDs and lifecycle. - [Claude Managed Agents](https://cwork101.com/platform/managed-agents): Server-hosted agents with a managed sandbox, billed on both tokens and session runtime rather than tokens alone. - [Claude Console](https://cwork101.com/platform/claude-console): The web interface for the Developer Platform — API keys, usage, spend, rate limits, and a workbench for testing prompts. ## Academy Anthropic Academy courses and the AI Fluency framework, mapped into learning paths with prerequisites and the wiki pages that go deeper. - [Anthropic Academy](https://cwork101.com/learn/anthropic-academy): Anthropic's official learning site at academy.claude.com — courses, tutorials and product learning paths, free to take. - [Claude 101 (Anthropic Academy course)](https://cwork101.com/learn/claude-101): Anthropic's official introductory course on Claude — 13 lessons and a quiz, covering what Claude is and how to work with it. - [AI Fluency: Framework & Foundations](https://cwork101.com/learn/ai-fluency-framework): Anthropic Academy's flagship course, teaching the 4D framework — Delegation, Description, Discernment, Diligence — for working effectively with AI. - [Claude Code 101](https://cwork101.com/learn/claude-code-101): Anthropic Academy's introduction to Claude Code — 12 lessons and a quiz, and the shortest of the four entry-level courses. - [AI Capabilities and Limitations](https://cwork101.com/learn/ai-capabilities-and-limitations): An Anthropic Academy course on how large language models actually work — next-token prediction, knowledge, working memory, steerability and context limits. - [Introduction to Claude Cowork](https://cwork101.com/learn/introduction-to-claude-cowork): Anthropic Academy's course on Cowork — 14 lessons and a quiz, the longest of the four entry-level courses. - [The 4 Properties of AI](https://cwork101.com/learn/four-properties-of-ai): A seven-minute Anthropic Academy tutorial on what makes AI capable in some situations and limited in others. - [Claude Platform 101](https://cwork101.com/learn/claude-platform-101): Anthropic Academy's introduction to the Developer Platform — 13 lessons and a quiz, and the on-ramp to the 67-lesson API course. - [Introduction to Model Context Protocol](https://cwork101.com/learn/introduction-to-mcp): Anthropic Academy's MCP course — 10 lessons and a quiz, with a separate advanced course covering the harder material. - [Building with the Claude API](https://cwork101.com/learn/building-with-the-claude-api): Anthropic Academy's flagship developer course — 67 lessons, 8 quizzes and 9 hours covering prompting, tool use, RAG, agents, MCP and production patterns. ## Plans & Limits Subscription plans, usage limits, rate limits and regional availability, with a single matrix showing which features each plan actually unlocks. - [Free](https://cwork101.com/plans/free): The no-cost Claude plan — full access to the assistant with capped usage, and no Claude Code, Cowork, Design or Science. - [Pro](https://cwork101.com/plans/pro): The entry paid plan, and the point at which Claude Code, Cowork, Design and Science become available. - [Max](https://cwork101.com/plans/max): The high-usage individual plan, sold in 5x and 20x multiples of Pro's usage, with higher output limits and priority access. - [Team](https://cwork101.com/plans/team): The business plan — central billing, SSO and admin controls, sold in standard and premium seats. - [Enterprise](https://cwork101.com/plans/enterprise): The compliance-grade plan — SCIM, audit logs, role-based access, spend limits, network controls and a HIPAA option — available self-serve or sales-assisted. - [API billing](https://cwork101.com/plans/api): Pay-per-token billing on the Developer Platform — entirely separate from Claude app subscriptions, with its own limits, its own console and its own bill. ## Safety & Policy How Anthropic governs Claude — Constitutional AI, the Claude Constitution, Responsible Scaling Policy, usage policy, system cards and transparency reporting. - [Usage Policy](https://cwork101.com/safety/usage-policy): Anthropic's rules for what Claude may be used for — the document your deployment is measured against, regardless of which plan or cloud you are on. - [Constitutional AI](https://cwork101.com/safety/constitutional-ai): The training method behind Claude's behaviour — a written set of principles the model is trained against, rather than behaviour shaped only by human preference labels. - [Responsible Scaling Policy](https://cwork101.com/safety/responsible-scaling-policy): Anthropic's framework for gating model capability behind safety levels, with defined thresholds and required safeguards at each. - [System cards and model cards](https://cwork101.com/safety/system-cards): Per-model documents reporting what evaluation found before release — capabilities, risks, safety testing and known limitations. - [Data usage and retention](https://cwork101.com/safety/data-usage): What Anthropic does with the data you send — how it varies by plan and platform, and where zero-retention options exist. ## Comparisons - [The API vs a Claude subscription](https://cwork101.com/compare/api-vs-claude-app): A Claude subscription is for a person using Claude. The API is for software calling Claude. They do not share limits, credits or accounts. If you need both, you buy both — and plenty of people do. - [Bedrock vs Google Cloud vs Foundry vs the Claude API](https://cwork101.com/compare/bedrock-vs-vertex-vs-foundry): Use the first-party Claude API unless procurement forces otherwise. If it does, prefer an Anthropic-operated route — Microsoft Foundry or Claude Platform on AWS — because those follow Anthropic's deprecation schedule. Bedrock and Google Cloud are partner-operated and set their own. - [CLAUDE.md vs auto memory](https://cwork101.com/compare/claude-md-vs-memory): CLAUDE.md is for what the team needs Claude to know: standards, intent, commands, conventions. Auto memory is for what Claude works out along the way. If it matters beyond your machine, it belongs in CLAUDE.md and in version control. - [Claude vs Claude Code vs Cowork](https://cwork101.com/compare/claude-vs-claude-code-vs-cowork): Claude is for thinking with; Claude Code is for changing a codebase; Cowork is for producing a deliverable from your files and connected apps. If the output is a conversation, use Claude. If it is a commit, use Claude Code. If it is a document, deck or spreadsheet, use Cowork. - [Hooks vs Agent Skills](https://cwork101.com/compare/hooks-vs-skills): If it must happen every time, it is a hook. If it needs judgement about when and how, it is a skill. Trying to guarantee behaviour with instructions is the most common configuration mistake in Claude Code. - [MCP vs Connectors](https://cwork101.com/compare/mcp-vs-connectors): If the person connecting it will never open a config file, you want a connector. If it is a developer wiring Claude Code or an SDK agent into an internal system, you want MCP. Many connectors are MCP underneath — the difference is who operates them, not what they are. - [Fable vs Opus vs Sonnet vs Haiku](https://cwork101.com/compare/opus-vs-sonnet-vs-haiku): Start with Opus 5. Move down to Sonnet 5 if it holds up on your evaluation — it is 60% cheaper and usually does. Use Haiku 4.5 for high-volume, latency-sensitive work that does not need a 1M-token window. Reach for Fable 5 only when a long-running agent is failing on capability, not on prompting. - [Plugins vs Agent Skills](https://cwork101.com/compare/plugins-vs-skills): Write a Skill when you are adding a capability. Publish a plugin when you are distributing one or more capabilities to other people — a plugin can contain skills, commands, hooks and MCP server configuration in a single installable unit. - [Pro vs Max](https://cwork101.com/compare/pro-vs-max): Max unlocks nothing that Pro does not already have. You are buying 5x or 20x the usage, higher output limits, early feature access and priority traffic. If you are not hitting Pro's limits mid-task, you do not need Max yet. - [Projects vs Agent Skills](https://cwork101.com/compare/projects-vs-skills): Use a Project when the same body of material comes up repeatedly — a client, a codebase, a research topic. Use a Skill when the same procedure comes up across different material. Material versus method is the whole distinction. - [Agent Skills vs MCP](https://cwork101.com/compare/skills-vs-mcp): A Skill is knowledge — instructions and scripts Claude loads when the task calls for them. An MCP server is a connection — access to a tool or data source. If your problem is "Claude does not know our process", write a Skill. If it is "Claude cannot reach our system", stand up an MCP server. - [Subagents vs Agent Teams](https://cwork101.com/compare/subagents-vs-agent-teams): Use a subagent when you want an answer without the context cost of finding it. Use an agent team when you have several genuinely parallel workstreams that each need their own persistent session — and ideally their own working directory. - [Team vs Enterprise](https://cwork101.com/compare/team-vs-enterprise): If your requirement is "one invoice, SSO, and our data is not used for training", Team covers it. If anyone has asked about audit logs, SCIM, role-based access, data retention or IP allowlisting, you need Enterprise — those are not add-ons to Team. ## Ecosystem directory - [Agent Skills (agentskills.io)](https://cwork101.com/ecosystem/agentskills-io): The open standard site for the Agent Skills format — specification, quickstart, and a showcase of the agents outside Anthropic that support it. - [Anthropic Discord](https://cwork101.com/ecosystem/anthropic-discord): The official community server, and the place a problem gets discussed before it reaches documentation — which is exactly why anything from it is a Tier 4 report. - [anthropics/claude-code](https://cwork101.com/ecosystem/anthropics-claude-code): The Claude Code repository — source, issue tracker, and the fastest route to reporting a bug or checking whether one is already known. - [anthropics/skills](https://cwork101.com/ecosystem/anthropics-skills): Anthropic's public repository of Agent Skills — the reference examples, the format specification, a skill template, and the document skills that power Claude's file capabilities. - [Claude Code documentation index (llms.txt)](https://cwork101.com/ecosystem/claude-code-docs-index): A machine-readable index of every Claude Code documentation page — the fastest way to find out whether a feature is documented at all. - [Claude Code plugin marketplaces](https://cwork101.com/ecosystem/claude-code-plugin-marketplaces): The distribution layer for Claude Code plugins — anyone can publish a marketplace, and organisations can steer which plugins their developers see. - [Claude Cookbooks](https://cwork101.com/ecosystem/claude-cookbooks): Anthropic's collection of 150+ working recipes — agent patterns, tool use, RAG, evals, caching and cost optimisation — as runnable notebooks rather than prose. - [MCP reference servers](https://cwork101.com/ecosystem/mcp-reference-servers): The official collection of Model Context Protocol reference implementations, plus pointers to community-built servers. - [modelcontextprotocol.io](https://cwork101.com/ecosystem/modelcontextprotocol-io): The Model Context Protocol's own site — the specification, the SDKs, and the client and server documentation that sit above any one vendor's implementation. - [Agents outside Anthropic that support Agent Skills](https://cwork101.com/ecosystem/skills-compatible-agents): Community project. Third-party coding agents and assistants that have adopted the Agent Skills format, making a skill you write portable beyond Claude. ## About - [About CWork101](https://cwork101.com/about): What this reference is, who runs it, and what it is not. - [Editorial policy](https://cwork101.com/editorial-policy): How pages are written, verified and corrected. - [How we source](https://cwork101.com/sources): The four source tiers and what each one licenses this reference to claim.