Reference
Claude Code feature map
Everything you’ll meet in Claude Code: how stable each feature is in the docs’ own words, and which plans and providers have it. Where the course teaches a feature, there’s a link to the lesson.
Updated
Where you run it
Every provider
its settings path reads “Claude Code [Beta]”
Every provider
beta on Linux only
A Claude subscription; on Bedrock, Agent Platform, Foundry or a Console API key through Claude Desktop set up for that provider
Pro, Max and Team; Enterprise with a premium or Chat + Claude Code seat
A Claude subscription
Pro and Max; Team and Enterprise once an admin turns it on
Pro and Max, in a Slack workspace not connected to Claude Tag; on Team and Enterprise, Claude Tag takes its place
Team and Enterprise, on Anthropic’s own service
Pro, Max, Team or Enterprise
on macOS
Pro and Max
Pro and Max
A Claude subscription
In a session
A Claude subscription, the Console and Claude Platform on AWS; on Bedrock, Agent Platform and Foundry with newer models only. Team and Enterprise admins can turn it off
Every provider
Every provider
Every provider
Every provider
A Claude subscription, paid for with usage credits (an Owner turns it on for Team and Enterprise), and Console organizations with access provisioned; not Claude Platform on AWS, Bedrock, Agent Platform or Foundry
requires the Anthropic API
A Claude subscription and the Console
A Claude subscription, the Console, Claude Platform on AWS and Foundry; not Bedrock; on Agent Platform with newer models only
Extending Claude Code
As for hooks
Every provider
A Claude subscription (Team and Enterprise once an admin turns them on) and the Console; not Claude Platform on AWS, Bedrock, Agent Platform or Foundry
Every provider
Parallel and unattended work
Every provider; on Pro, turn them on in /config
Every provider
Every provider
A Claude subscription; elsewhere between sessions on the same machine only
public beta, rolling out gradually
Pro and Max; not yet Team or Enterprise
Every provider
Pro, Max, Team and Enterprise
Every provider
A Claude subscription, the Console, Bedrock, Agent Platform and Foundry; not Claude Platform on AWS
Every provider except Microsoft Foundry
Review and security
Team and Enterprise
A Claude subscription
All plans; it needs Python 3.7 or later, and 3.10 or later for the agentic commit review and, on a third-party provider, for every model-backed review
A paid plan, Anthropic API access or a third-party provider
Other
Pro, Max, Team and Enterprise, on the Anthropic API
On by default on Pro, Max and Team; on Enterprise an Owner turns each template on
Model aliases
Pick one with /model, claude --model or the model setting. An alias moves to a newer model over time; default is not a model but a reset of your choice.
| sonnet | the latest Sonnet | everyday coding work |
| opus | the latest Opus | complex reasoning |
| fable | your provider’s Fable model | the hardest and longest tasks |
| best | Fable if your account has it, otherwise Opus | the strongest model available to you |
| haiku | Haiku | quick, simple tasks |
| opusplan | Opus in plan mode, then Sonnet carries out the plan | deep planning, cheaper execution |
| opus[1m] | Opus with a one-million-token window | very long sessions where opus has no such window |
Effort levels
| low | quick back-and-forth where you check every result: a sketch, a rename |
| medium | everyday work with clear boundaries, such as a new function |
| high | work where edge cases are likely: fixing a bug in existing code |
| xhigh | deeper reasoning at a higher token cost |
| max | hard problems Claude solves without you; check before relying on it — it can overthink |
Which setting overrides which, and the traps in max and ultrathink — the lesson “Models, effort and fast mode”.
Checked against the Feature availability page and each feature’s own page for Claude Code v2.1.306, 2026-10-22. Prices and limits are on claude.com/pricing.
Written by Mikhail Kuzmitskii