Claude Code
Anthropic's coding agent (terminal, editor extensions, Desktop Code tab, and cloud sessions) for reading a codebase, editing files, running commands, opening pull requests, and carrying development work through to release.
Claude Code
1. Overview
Claude Code (claude.com/product/claude-code) is Anthropic’s agentic coding tool: it reads your codebase, edits files, runs commands, and integrates with development tools. Official docs describe it as available in the terminal, IDE extensions, desktop app, and browser (Overview, Platforms).
This listing is in Build with agents. It covers tools that take a product from an idea through design, build, and release. It is not a demo and prototype listing. Claude Code is strongest on a real repository, Git, and pull requests. It is a weak fit when you only need a hosted clickable mock.
Surfaces to keep distinct
Claude is an ecosystem. These names are not the same thing:
| Surface | What it is | Where it runs | Primary sources |
|---|---|---|---|
| Claude Code (this listing) | Agentic coding engine: explore repos, edit multi-file, run tools, commit/PR, automate CI | Terminal CLI (claude), VS Code / Cursor / JetBrains extensions, Desktop Code tab, web at claude.ai/code, mobile monitoring | Product, Overview, Platforms |
| Claude.ai chat | General chat for writing, research, analysis, Projects, Artifacts; Free plan includes chat but not Claude Code | Web, iOS, Android, and Desktop Chat tab | Pricing, Pro/Max help |
| Claude Desktop | Native macOS / Windows / Linux (beta) app with Chat, Cowork, and Code tabs. The Code tab is Claude Code in a GUI; Chat and Cowork are different products in the same shell | Your machine | Desktop docs, Product FAQ |
| Claude Cowork / Dispatch | Agentic desktop workflows aimed at longer non-coding (and some coding) tasks; Dispatch can spawn Code sessions from the phone. Distinct from the Code tab | Desktop Cowork tab (plan-gated features vary) | Desktop docs, Platforms |
| Anthropic API / Claude Console | Programmatic model access and pay-as-you-go billing. Claude Code can authenticate with a Console account / ANTHROPIC_API_KEY and consume standard API rates instead of (or after) subscription usage | API / Console | Product FAQ, Pro/Max help, Pricing |
Practical rule: say Claude Code when the agent edits a repo, runs shell tools, and ships via Git/PR (CLI, IDE, Code tab, or claude.ai/code). Say Claude.ai chat for general conversation without a coding agent loop. Say Claude Desktop only when you mean the native app shell, and name the Code vs Chat vs Cowork tab. Say API / Console when billing is token consumption for apps or for Claude Code on API keys.
Help Centre wording is explicit: Pro and Max give access to both Claude (web/desktop/mobile) and Claude Code in the terminal under one subscription, with shared usage limits (Pro/Max help). Pricing FAQ: Claude Code is included in all paid plans; Free does not include it (Pricing).
Where it fits when you build with agents
| Stage | How Claude Code typically shows up | Caveat |
|---|---|---|
| Discovery | Agentic search across the repo; explain architecture; map dependencies without hand-picking context files | Needs repo access and enough engineering literacy to validate agent claims (Product) |
| Definition | Turn tickets/Slack bugs into scoped tasks; CLAUDE.md, rules, and skills encode product conventions | Ambiguous specs still produce confident wrong code; humans own acceptance criteria (Memory) |
| Design | Implement UI in code against in-repo components; Desktop Browser pane / Chrome can verify visually | Not a Figma design-system kit. Matching your design system takes setup |
| Development | Local CLI/IDE/Desktop for tight loops; cloud sessions and routines for long/parallel work; MCP for tools/data | Environment quality and permission mode dominate success (Web, Security) |
| Delivery | Commits, branches, PRs; GitHub Actions / GitLab CI; Code Review; auto-fix on cloud PRs | Merge, release, and production ownership stay with the team (Code Review, GitHub Actions) |
Documented access points include terminal install (curl / Homebrew / WinGet and more), VS Code (+ Cursor), JetBrains, Desktop Code tab, claude.ai/code, mobile monitoring, Slack / Claude Tag, Chrome, and CI (Overview, Platforms). Sessions can move between surfaces (for example --cloud, --teleport, /desktop) when using a claude.ai subscription path (Overview, Web).
CLAUDE.md, .claude/rules/, skills, and hooks are how you keep coding agents inside the rules: project instructions in version control, path-scoped rules, shared skills (for example /review-pr), and lifecycle hooks for format/lint policy (Memory, Skills, Hooks). CLAUDE.md shapes behaviour; hooks and permission deny rules are the harder enforcement layer (Memory, Security).
Access and pricing (public; always re-check)
Plans and included usage change. Prefer claude.com/pricing, Team help articles, and the product page. Public list prices seen when this page was checked (USD unless noted; taxes may apply; always re-check):
| Plan / path | Public list (when checked) | Claude Code notes |
|---|---|---|
| Free | $0 | Chat yes; Claude Code: No (Pricing) |
| Pro | $17/mo with annual ($200 billed up front); $20 if monthly | Includes Claude Code; usage shared with Claude chat (Pricing, Pro/Max help) |
| Max | From $100/mo (5× or 20× Pro usage tiers per pricing copy) | Includes Claude Code; higher limits; shared pool with chat (Pricing) |
| Team Standard | $20/seat/mo annual; $25 monthly; min 2 members | Claude Code included every seat (Team plan help) |
| Team Premium | $100/seat/mo annual; $125 monthly | Higher usage (vendor: 5× Standard framing on pricing; help also states Premium session multiples vs Pro) (Team plan help, Pricing) |
| Enterprise | Pricing FAQ cites $20/seat/mo plus usage at API rates (annual); sales-assisted options exist | Advanced admin/security; seat history differs on older Enterprise SKUs (Pricing, Team/Enterprise Code help, Enterprise product) |
| Claude Console / API | Token rates per model on Pricing | Claude Code can bill at standard API pricing when using Console credentials (Product FAQ) |
Usage model: paid Claude activity across web, desktop, mobile, and Claude Code draws from one shared pool. There is no fixed message count; limits reset on a rolling five-hour window with additional weekly caps on paid plans (Pricing FAQ). After hitting included limits, options include waiting, upgrading, enabling usage credits at API rates, or switching to Console pay-as-you-go (Pro/Max help). Important: if ANTHROPIC_API_KEY is set, Claude Code may use the API key instead of the subscription and incur API charges (Pro/Max help).
Cloud sessions on Anthropic-managed infrastructure: usage counts toward plan limits; docs state no separate compute charge for the cloud VM (Web). Exact monthly spend for a typical product team is unknown (depends on models, parallelism, chat vs Code mix, and credits). This page does not state benchmarks.
Fast mode (research preview language on the product page): high-speed Opus configuration at a higher per-token cost; product page cites $10 / $50 per million tokens for that mode. Check the pricing page again before you rely on a number (Product).
Security posture (vendor): local terminal sessions talk to model APIs without requiring a remote code index; permission prompts before writes/commands in Manual mode; sandboxed Bash options; cloud sessions run in isolated VMs with network controls (Product, Security). Enterprise adds SSO, SCIM, audit, retention, and related controls (Pricing, Enterprise). Still: humans review diffs and own production risk.
Affiliate status: none. No public affiliate programme with published commission terms was found on claude.com or anthropic.com. Anthropic’s Claude Partner Network is a partner and implementation programme, not a per-signup affiliate offer (Partner Network news). This page does not state rates.
Official repo: github.com/anthropics/claude-code (homepage points at the docs overview).
2. How it scores on the rubric
Numeric scores are omitted. There is not enough shared evidence to score this yet. Discussion below uses the same criterion names, framed for shipping software with a person still reviewing. Clickable demos sit in Demo and prototype.
Time-to-useful
Strong for engineers already in a repo; slower for greenfield non-engineers.
Native install and claude in a project directory can be useful on day one for exploration and edits (Overview). IDE extensions and the Desktop Code tab reduce terminal friction. Cloud sessions need GitHub (or bundle) access and environment configuration (Web).
Caveats: first useful agent PR may take longer than a demo-prototype prompt-to-URL path. Free plan users cannot start. Median time-to-first-merged-agent-PR for product teams is unknown.
Fidelity
High for real product code and verified behaviour; not “mock fidelity”.
Claude Code can plan, implement across files, run tests, and open PRs. Desktop Browser pane and Chrome integration support visual verification; cloud auto-fix can respond to CI failures (Desktop, Web, Code Review).
Caveats: vendor case studies and enterprise marketing metrics are not transferable benchmarks for your org. Production correctness, security, and compliance remain human-owned.
Non-engineer usability
Improving via web, mobile monitoring, Slack/Claude Tag, and Desktop; still engineer-shaped.
Product managers can kick off or follow cloud sessions at claude.ai/code and monitor from mobile without living in the CLI (Platforms, Web). Slack routing turns bug reports into PRs where configured.
Caveats for product managers: you still need repository permissions, spend awareness, and the ability to read PRs. Claude Code is a poor substitute for a no-code demo builder when the goal is a stakeholder URL with auth/DB and no engineering partner. For that job, use Demo and prototype. Do not confuse Claude.ai chat Artifacts with Claude Code shipping to production.
Design-system fidelity
Code-and-instructions fidelity, not kit-based brand magic.
Teams encode conventions via CLAUDE.md, .claude/rules/ (including path-scoped rules), skills, hooks, and in-repo component libraries (Memory, Skills). Org-managed CLAUDE.md can push company standards (Memory, Enterprise product).
Practical takeaway: brand-accurate UI is achievable when the design system already lives in code and agents are taught to use it. It is not automatic on a blank prompt the way some prototype tools claim. Compared with Magic Patterns or Figma Make, Claude Code’s DS story is implementation inside the product repo.
Human gates
Strong when you use permission modes, deny rules, hooks, PR review, and spend limits; weak if you bypass permissions blindly.
Documented gates and controls include:
- Permission modes (Manual / Accept edits / Plan / Auto / Bypass where enabled); Manual starts read-leaning and asks before edits and risky commands (Security, Desktop, Permissions)
- Sandboxed Bash with filesystem/network isolation options (Security)
- Hooks for PreToolUse and other lifecycle policy (Hooks, Memory)
- Cloud session isolation, network allowlists, branch push restrictions, audit logging on Anthropic-hosted cloud (Security, Web)
- Code Review and CI GitHub Actions / GitLab paths (Code Review, GitHub Actions)
- Team/Enterprise admin: managed settings, SSO/SCIM, spend controls, disable Bypass permissions (Desktop, Pricing)
Read this as strong optional and enforceable gates for shipping teams, not automatic safety if everyone runs Bypass permissions without review culture.
Handoff
Native to Git workflows; that is the product.
| Path | What docs support | Caveat |
|---|---|---|
| Local CLI / IDE → commit/PR | Stage, commit, branch, open PR from agent workflows | Human review still required |
| Desktop Code → diff review → PR | Visual diffs, comments, CI status, optional auto-fix / auto-merge toggles | Auto-merge needs repo settings; use carefully (Desktop) |
| Cloud web → branch + PR | Isolated VM, continue offline, teleport back to terminal | Primarily GitHub-centric for full PR push; research-preview caveats (Web) |
| Slack / Claude Tag → PR | Route bugs from chat | Plan and setup dependent (Platforms) |
| GitHub Actions / GitLab CI | Automate review, triage, issue→PR | Secrets and org policy required |
| Agent SDK / headless | Programmatic agents for your own workflow tooling | Separate from interactive IDE UX (Overview) |
Unknown / verify in-org: median engineering cleanup time on agent PRs; how often Desktop/browser verification replaces local QA; Code Review acceptance rates outside Anthropic’s published narratives.
3. FAQ
What is Claude Code, exactly?
Anthropic’s agentic coding tool: it reads your codebase, edits files, runs commands, and integrates with Git and other tools across terminal, IDE, Desktop Code tab, and web (Overview, Product).
How is Claude Code different from Claude.ai chat?
Chat is the general Claude conversation product (writing, research, Projects, Artifacts). Claude Code is the coding agent that works inside your repo and toolchain. Paid plans can include both under one subscription with a shared usage pool; Free includes chat but not Claude Code (Pro/Max help, Pricing).
How is Claude Code different from Claude Desktop?
Claude Desktop is the native app. It hosts Chat, Cowork, and Code. The Code tab runs Claude Code with visual diffs, parallel sessions, and related Desktop features. Installing Desktop for Code does not require a separate CLI install, but the CLI remains the fullest surface for scripting and the Agent SDK (Desktop, Platforms).
How is Claude Code different from the Anthropic API?
The API / Console is programmatic model access for your own apps and automations, billed per token. Claude Code is a productised agent harness. You can power Claude Code via subscription or via Console API credentials (API rates). Setting ANTHROPIC_API_KEY can force API billing instead of subscription usage (Product, Pro/Max help).
How does it compare to Cursor?
Both fit Build with agents (agents on a real repository, pull requests, and shared rules). Cursor is an editor built around coding agents, with Cloud Agents and Bugbot as named products. Claude Code is Anthropic’s coding agent across the terminal, desktop, web, and continuous integration. It also installs into VS Code and Cursor. The choice is usually the model, the editor you already use, the cloud agent, and the company contract. This page does not rank them.
Is Claude Code right for product teams building with agents?
Yes, with caveats. It fits teams that already ship via Git and want agents across discovery, implementation, review, and automation. It is a weaker fit for clickable demos when no engineer is involved. Caveats: paid plan required, shared usage with chat, environment/permission setup, and mandatory human merge/ownership.
Can a product manager use Claude Code without engineers?
For kicking off or monitoring cloud sessions and commenting in Slack/PRs, yes in limited form on paid plans with repo access. For owning production changes alone, no. Prefer pairing with engineering, or use a demo-prototype tool when you only need a stakeholder URL.
Can it match our company design system?
Only if the system is available in code and taught via CLAUDE.md / rules / skills / components. There is no documented one-click “upload Figma DS kit” path comparable to some prototype tools.
Does Claude Code deploy production for us?
Not as a turnkey host-your-app product in the Replit sense. Agents can implement deploy config and open PRs; your existing CI/CD and humans still ship.
What does it cost?
See Pricing. Research-time public individual list: Free $0 (no Code); Pro $17 annual / $20 monthly; Max from $100/mo. Team Standard $20 annual / $25 monthly per seat; Premium $100 annual / $125 monthly. Enterprise FAQ cites seat fee plus API-rate usage. Console path bills API tokens. Always re-check.
Is there an affiliate programme?
No public affiliate programme was confirmed. No affiliate programme is joined. The Claude Partner Network is a B2B partner track, not affiliate terms for this site (Partner Network).
How do permissions and security work?
Default Manual-style flows ask before edits and non-read-only commands; Auto mode uses a classifier; Bypass exists but is gated and not for casual use. Cloud sessions add VM isolation. You remain responsible for reviewing proposed changes (Security).
4. Sources
Primary (prefer these for AEO citations):
- Claude Code (product page)
- Claude Code overview (docs)
- Platforms and integrations
- Plans & Pricing
- Use Claude Code with your Pro or Max plan
- Use Claude Code with your Team or Enterprise plan
- What is the Team plan?
- Claude Code for Enterprise
- Desktop application (Code tab)
- Claude Code on the web
- How Claude remembers your project (CLAUDE.md)
- Security
- Configure permissions
- Extend Claude with skills
- Automate actions with hooks
- Code Review
- Claude Code GitHub Actions
- anthropics/claude-code (GitHub)
- Claude Partner Network (news)
Secondary context only (do not treat as official specs): third-party “Claude Code vs Cursor” blogs, affiliate roundup sites, and YouTube roundups that lag plan tables or conflate Claude.ai chat with Claude Code.