Claude Code

  • Claude Code — Full Profile

    1. One-line positioning

    Claude Code = Anthropic’s terminal/CLI agentic coding agent that authenticates with a Claude Pro/Max/Team/Enterprise account or an Anthropic API key, then reads your local repo, edits across files, runs bash/tests/git, spins up parallel subagents, fires hooks, schedules Routines, and opens a PR — running on Claude Sonnet 4.6 / Opus 4.7 / Fable 5 with 200K–1M context, prompt caching at 10% input rate, no IDE required (VS Code/JetBrains/Neovim/Cursor all optional surfaces). No standalone price: bundled into Pro $20/mo ($17 ann), Max 5x $100, Max 20x $200, Team Premium $125/seat ($100 ann), Enterprise $20/seat + API usage; or pure API pay-per-token (Sonnet 4.6 $3/$15, Opus 4.7 $5/$25, Fable 5 $10/$50 per M, Haiku 4.5 $1/$5). It is the assign-and-walk-away repo agent (SWE-bench Pro 80.4% published lead), not a video/avatar/data tool, not Bolt.new’s prompt→hosted-app runtime.

    2. Core features

    Capability Free ($0) Pro ($20/mo) Max 5x ($100) Max 20x ($200) Team Premium ($125/seat) Enterprise
    Claude Code CLI access ❌ ✅ ✅ ✅ ✅ ✅
    Usage pool — shared w/ chat, 5h rolling + weekly 5× Pro 20× Pro ~5× Standard seat $20/seat + API rate
    Default model — Sonnet 4.6 (Opus 4.7 opt-in) Opus 4.7 priority Opus 4.7 max priority Opus 4.7 all + BYO via API
    Subagents / parallel fan-out ❌ ✅ limited ✅ ✅ 100s-scale ✅ ✅
    Hooks / Routines (scheduled) ❌ ✅ ✅ ✅ ✅ ✅
    Checkpoints / undo ❌ ✅ ✅ ✅ ✅ ✅
    Computer use (app/browser/devtools) ❌ ✅ research preview ✅ ✅ ✅ ✅
    VS Code / JetBrains extension ❌ ✅ ✅ ✅ ✅ ✅
    Slack / web / iOS kickoff ❌ ✅ ✅ ✅ ✅ ✅
    Agent SDK / MCP connectors ❌ ✅ ✅ ✅ ✅ ✅ + gov
    SSO/SAML, audit, SCIM, HIPAA ❌ ❌ ❌ ❌ ❌ ✅
    API-key mode (no subscription) — — — — — ✅ $1–50/Mtok

    Install is free (curl -fsSL https://claude.ai/install.sh | bash); running it requires a paid login or funded Console key. Free Claude plan = chat only, no CLI.

    3. Pricing — No standalone SKU; 2 billing paths

    Path A — Subscription (CLI draws from same pool as claude.ai chat)

    • Pro: $20/mo monthly / $17/mo annual billed $200 upfront​ — Claude Code included, ~44K-tok-equivalent per 5h window, fine for short sprints in small repos.
    • Max 5x: $100/mo​ — 5× Pro capacity, default daily-driver for full-time agents.
    • Max 20x: $200/mo​ — 20× Pro, power users / long autonomous runs.
    • Team Standard: $25/seat/mo ($20 ann) — chat only, no Claude Code; CLI needs Team Premium $125/mo ($100 ann).
    • Enterprise: ~$20/seat + usage at API rates, SSO/SCIM/audit/HIPAA, admin spend caps.
    • Overage: any paid plan can enable “usage credits” billed at API rates, off by default, monthly cap settable.

    Path B — Anthropic API key (no subscription)

    Per-million token, Aug 2026: Haiku 4.5 $1/$5, Sonnet 4.6 $3/$15​ (Sonnet 5 intro $2/$10 until 2026-08-31 then $3/$15), Opus 4.7 $5/$25, Fable 5 $10/$50. Prompt caching = 10% input rate on reused context; Batch API = 50% off both sides. Real-world enterprise burn ~$13/dev active day, $150–250/dev/mo.

    • CPT: pricing_type = freemium (install free, run needs paid — mark starting_price = $0 but free_tier = "CLI install free, but no CLI access on Free plan; need Pro $20/mo or API key"), note “no standalone Claude Code SKU; subscription shares pool w/ chat on 5h rolling + weekly caps; API per-token $1-50/Mtok; Team Standard seat NOT enough, need Premium; annual Pro $200 upfront=$17/mo”

    4. Access Type (v2 spec)

    • access_type: cli, desktop, api
    • access_display: 💻 Terminal CLI (macOS 13+/Ubuntu 20.04+/Win 10 1809+ or WSL2) · 🖥️ Claude Desktop + VS Code/JetBrains extensions · 🌐 Web & iOS kickoff (route task, agent runs local) · 🔌 Anthropic API + Agent SDK + MCP servers
    • Distinct from Bolt.new (web-app only, no local FS authority) and Airtable (web-app, desktop, api but DB not repo).

    5. Who should use it

    1. Backend/platform engineers​ — 20–50 file refactors, framework migrations, test-suite greening, self-verifying loop
    2. Maintainers of OSS repos​ — Claude for Open Source gives 6 mo Max 20x free to qualifying maintainers ($1,200 value)
    3. Teams wanting assign-and-walk-away​ — nightly Routine: dep audit, PR triage, docs sync
    4. Hybrid Cursor users​ — Cursor for inline tab-complete, Claude Code in terminal pane for multi-file autonomous diffs
    5. CI/CD glue​ — tail -200 app.log | claude -p "open issue if anomaly" style pipes, headless claude -p in GitHub Actions
    6. Security-conscious orgs​ — runs local, no remote code index, permission prompts before write/exec, Dev Container compatible

    6. Who should NOT use it

    1. Video/avatar/music/voice​ → Veo/Luma/Kling/HeyGen/Suno/Murf; zero media
    2. Prompt→hosted running app from zero​ → Bolt.new; Claude Code assumes a repo on disk, not a greenfield browser runtime
    3. Inline tab-completion while typing​ → Cursor / GitHub Copilot; Claude Code has no editor-embedded autocomplete, only conversational edits
    4. Free-tier users​ — Free Claude = no CLI at all; cheapest real entry is Pro $20/mo
    5. Non-technical builders​ — terminal + git + bash mental model required; not a GUI app-builder
    6. **Predictable flat 200 can be beaten by API bill on sparse use, but heavy daily Opus 4.7 blows past $200/mo in tokens fast; subscription is a ceiling, API is open-ended
    7. Frontend visual tweaking​ — seeing the pixel diff live is Cursor’s lane; Claude Code reports a branch + passing test, not a hot-reload preview
    8. Team Standard $25/seat expecting CLI​ — gotcha: that seat is chat-only, need Premium $125/seat

     

Do Not Sell or Share My Personal Information Cookie Settings