Which Claude product should you use?
Lightbridge Labs compares the Claude product family so teams pick the right surface for the work. The lineup spans claude.ai for chat, Claude Cowork for agentic file work, Claude Code for terminal coding, the Claude API for programmatic access, Claude Design for visuals, Claude Security for code scanning, and Claude on major cloud providers.
The Claude product family at a glance: seven ways to use Claude.
The Anthropic product surface has grown well beyond the chat interface most people start with. Each surface targets a different user and a different shape of work. This table from Lightbridge Labs maps every Claude product to where it runs, who it is for, and what it does best.
| Product | Where it runs | Who it is for | Best for | Autonomy |
|---|---|---|---|---|
| Claude (claude.ai) | Browser, mobile, and desktop apps | Everyone, no technical setup | Chat, drafting, summarizing, and questions on content you paste in | One conversation, one prompt at a time |
| Claude Cowork | Claude desktop app | Non-technical knowledge workers | Multi-step file work: organize, assemble, and extract across a folder | Completes goals end to end |
| Claude Code | Terminal command line | Software developers | Writing code, running tests, and managing git in a project | Full session autonomy across the repository |
| Claude API and Anthropic SDK | Your own application code | Developers building on Claude | Programmatic access to build applications powered by Claude | You control the loop |
| Claude Design | Within the Claude interface | Founders, product managers, non-designers | Quick visuals: prototypes, slides, and one-pagers | Single-shot generation (research preview) |
| Claude Security | Connected code repositories | Enterprise security teams | Scanning code for vulnerabilities and proposing fixes | Per-finding review (public beta) |
| Claude on cloud providers | Amazon Bedrock, Google Vertex AI, Microsoft Foundry | Enterprise infrastructure teams | The same models as the API, billed through your existing cloud account | Same as the API |
Note: product surfaces and model availability change quickly. This guide describes tiers and capabilities qualitatively rather than quoting prices or usage limits, which go stale fast. Verify current details at docs.anthropic.com and anthropic.com.
A decision tree for choosing the right Claude product for the task.
Start from the work, not the product. Match the sentence that fits the job in front of you, and the right Claude surface follows. These choices are not mutually exclusive: most teams use several together.
I want to ask a question or get help drafting something.
I have a folder of files and need a finished deliverable from them.
Use Claude Cowork
I am writing or editing code inside a project.
Use Claude Code
I am building an application that calls Claude programmatically.
I need a quick visual mockup, slide deck, or one-pager.
Use Claude Design
I want a scan of my codebase for security issues, with proposed fixes.
Use Claude Security
My company already runs on AWS, Google Cloud, or Azure and wants Claude through that account.
Claude (claude.ai) is the conversational interface for everyone.
Claude at claude.ai is the chat product most people mean when they say Claude. You type, Claude responds, and it is available in the browser, on iOS and Android, and as a desktop app with no setup. It is fast to start and strong at writing, summarizing, analyzing content you paste in, and answering questions. The limits are deliberate: it sees only what you paste, it does not access your files or run code, and by default it starts each conversation fresh with no memory of earlier sessions.
Plan tiers run from a free tier through Pro, Max, Team, and Enterprise, with higher tiers unlocking more capable models, longer context, more usage, and admin and security features. The skill that makes claude.ai pay off is the same one that transfers to every other surface here: clear prompting. See how to prompt Claude for practical guidance.
Claude Cowork is the agentic desktop product for knowledge work.
Claude Cowork is Anthropic's agentic desktop application. You set a goal, and Cowork reads the files in folders you grant it, moves between applications, and returns a finished deliverable. It is built for non-technical knowledge workers in research, analysis, operations, legal, and finance, where the bottleneck is assembling work across many files rather than answering a single question. Cowork acts only on folders you explicitly grant, surfaces judgment-heavy decisions for your review, and runs through the Claude desktop app on paid plans.
Typical jobs are organizing files, assembling a report from source material, synthesizing research, and extracting data from unstructured documents. For a deeper walkthrough of how it works and where it fits, read what is Claude Cowork.
Claude Code is the terminal coding agent for software developers.
Claude Code is Anthropic's command-line tool. It runs in the terminal as a coding partner with full access to a repository: it reads code, edits files directly, runs tests and commands, and manages git. Persistent project context lives in a CLAUDE.md file at the repository root, and the tool is composable through hooks and skills. It expects an installation, a terminal, and a paid Claude account, so it is built for users comfortable with command-line workflows.
Where Cowork works on documents, Claude Code works on code. For the full picture of what it does and how teams adopt it, read what is Claude Code.
The Claude API and Anthropic SDK are the programmatic foundation under everything.
The Claude API is the HTTP interface for invoking Claude models, with official SDKs for Python, TypeScript, and other languages. It is the foundation the other products call internally: claude.ai, Cowork, Claude Code, and Design all run on the API. It gives developers full control over tool use, streaming, structured output, vision, and multi-turn conversations, along with controls such as prompt caching and batch processing. In return, you own the application around it, including authentication, rate limiting, error handling, and prompt design, and cost accrues with usage.
Reach for the API when Claude must run inside software with no person typing each prompt. Turning that into a governed, production-grade capability, with scoped permissions, signed artifacts, and audit evidence, is the work of custom AI development at Lightbridge Labs.
Claude Design generates visuals from a description, as a research preview.
Claude Design is an experimental Anthropic Labs product for generating visuals from a text description: prototypes, slides, one-pagers, and marketing graphics, without graphic design skills. It is built for founders, product managers, and other non-designers who need to communicate an idea visually. Because it is a research preview, the feature set evolves and it is not intended for production-critical visual workflows or the fine-grained control of a dedicated design tool. It runs on Claude's most capable image-generation tier inside the Claude interface on paid plans.
Claude Security scans connected repositories and proposes fixes, in public beta.
Claude Security is an Anthropic enterprise product, in public beta, that scans connected code repositories for vulnerabilities and proposes fixes. It reasons about code semantics rather than matching patterns alone, so it can go beyond traditional scanners, and it combines scanning and fix proposals in one workflow. It is built for enterprise security teams managing code at scale and is currently scoped to Claude Enterprise customers, with capabilities and pricing still evolving while it remains in beta.
Claude runs on Amazon Bedrock, Google Vertex AI, and Microsoft Foundry for cloud-native teams.
Claude is available as a model selection inside Amazon Bedrock (billed through AWS), Google Cloud Vertex AI (billed through Google Cloud), and Microsoft Foundry (billed through Azure). The models and their behavior are the same as direct API access. What changes is the operating path: the cloud route helps organizations that already run on a provider with vendor consolidation, regional data residency, and integrated identity and access management. The trade-offs are that latency can differ and a new model may arrive on a cloud provider some time after it reaches the direct Anthropic API.
Deciding between the direct API and a cloud-provider path is an architecture and governance question as much as a billing one. Lightbridge Labs covers that trade-off as part of AI strategy.
Claude plan tiers unlock products as you move from individual to enterprise.
Anthropic structures plans from a free tier through Pro and Max for individuals, Team for small groups, and Enterprise for larger organizations. The free tier is for trying Claude. Paid individual tiers add more usage and access to products such as Cowork and Design. Team adds shared administration and billing. Enterprise adds single sign-on, audit logs, data-retention controls, and access to Claude Security. The exact usage limits and prices shift over time, so confirm what a given tier includes at anthropic.com rather than relying on a fixed figure.
If you work under an enterprise agreement or through a partner, your account team can confirm which tier you are on and which products it unlocks.
Most teams use several Claude products together, not one.
These products are not mutually exclusive, and the strongest setups mix them. A common pattern uses claude.ai for ad-hoc questions, Claude Cowork for recurring document workflows, Claude Code for any custom code, and the Claude API for anything embedded in a system. A finance team might use claude.ai for one-off questions and Cowork for batch work on a folder of documents. A development team might use Claude Code in the terminal and the API for production-side automation.
The skill that transfers across all of them is clear prompting: state one task per message, the context Claude needs, and the shape of the output you want. Lightbridge Labs helps organizations decide where each surface fits, then governs them together through AI strategy and, where production software is needed on these surfaces, custom AI development.
This guide is independent guidance from Lightbridge Labs. Claude, Claude Cowork, Claude Code, Claude Design, Claude Security, and Anthropic are trademarks of Anthropic. Amazon Bedrock, Google Vertex AI, and Microsoft Foundry are trademarks of their respective owners. Lightbridge is not affiliated with, partnered with, or endorsed by Anthropic, Amazon, Google, or Microsoft.
Which Claude product to use: frequently asked questions
- What is the difference between Claude Cowork and Claude Code?
- Both are agentic, meaning they complete multi-step work rather than answering a single prompt, but they target different users and different material. Claude Cowork runs in the Claude desktop app and works on files in folders you grant it, so it suits non-technical knowledge workers who need a deliverable assembled from documents. Claude Code runs in the terminal with full access to a code repository, so it suits software developers who need code written, tests run, and git managed. Choose Cowork for document and knowledge work, and Claude Code for software development.
- When should I use the Claude API instead of claude.ai?
- Use claude.ai when a person is in the loop for a one-off task: a question, a draft, or analysis of content you paste in. Use the Claude API when Claude needs to run inside software, with no human typing each prompt. The API gives you programmatic control over tool use, streaming, structured output, and multi-turn conversations, which is what you need to embed Claude in an application or automation. The trade-off is that you build and operate everything around the call, including authentication, error handling, and prompt design.
- What is the difference between the Claude API and Amazon Bedrock?
- The models are the same. The difference is the billing and operations path. Calling the Claude API directly bills through an Anthropic account and gives you the newest models first. Calling Claude through Amazon Bedrock, Google Vertex AI, or Microsoft Foundry runs the same Anthropic models but bills through your existing cloud provider, which helps with vendor consolidation, regional data residency, and integrated identity and access management. The trade-off is that a new model can arrive on a cloud provider some time after it reaches the direct API.
- Which Claude product is best for a non-technical team?
- For ad-hoc questions and drafting, claude.ai needs no setup and works in a browser or app. For repeatable work across a folder of documents, such as organizing files, assembling a report from source material, or extracting data from many unstructured files, Claude Cowork is the better fit because it acts autonomously on the folders you grant it. Neither requires writing code. Lightbridge Labs often pairs the two: claude.ai for one-off tasks and Cowork for the recurring document workflows.
- Do these Claude products share the same underlying models?
- Yes. claude.ai, Cowork, Claude Code, Claude Design, Claude Security, and the cloud-provider variants all run on the same family of Anthropic models, and the API is the foundation they call internally. As of mid-2026 that family includes Claude Fable 5 as the most capable model, Claude Opus 4.8, Claude Sonnet 4.6 as the balanced default for most work, and Claude Haiku 4.5 for fast, lightweight tasks. Product surfaces and model availability change quickly, so verify current details at docs.anthropic.com and anthropic.com.
- Can I use more than one Claude product at the same time?
- Yes, and most teams do. The products are not mutually exclusive. A common pattern uses claude.ai for ad-hoc questions, Claude Cowork for recurring document workflows, Claude Code for any custom code, and the Claude API for anything embedded in a system. The skill that transfers across all of them is clear prompting: state what you want, the context Claude needs, and the shape of the output. Lightbridge Labs helps organizations decide where each surface fits and how to govern them together.
- How does Lightbridge Labs help organizations choose between Claude products?
- Lightbridge Labs provides independent guidance on which Claude surface fits a given workflow, then helps put the chosen surface into production with the security, audit, and governance a regulated buyer expects. The advisory work begins with AI strategy, mapping use cases to the right tools. Where a team needs production software on these surfaces, such as custom skills, plugins, MCP connectors, and deterministic hooks, that is custom AI development. Lightbridge Labs is independent and is not affiliated with or endorsed by Anthropic or any cloud provider.
From choosing a Claude product to running it in production.
Lightbridge Labs maps the right Claude surface to your workflow, then helps put it into production with the security, audit, and governance a serious buyer expects.