1.1.49

Tapproval Base is the computer half of Tapproval: it puts Claude Code's permission prompts on your Apple Watch, each marked with its risk, and one tap answers. Nothing is approved for you unless you ask it to be, and a CRITICAL never is.

Already installed? Starting Claude Code once updates it.

Since 1.1.48

• A session that is quietly running a background task is no longer shown as idle. If Claude has handed work to a background task and is waiting for it, the session list says it is working.
• Groundwork for Codex, continued. If you also use OpenAI's Codex, the helper now answers only the questions Codex actually asks you. Each card is named after the command, for example "Fetch https://example.com". Codex is found inside the ChatGPT app, and until you have trusted the hook in Codex, the watch's Connection screen tells you how: open codex in Terminal once and choose "Trust all and continue". Nothing changes for Claude Code, and full Codex support on the watch comes in later releases.

1.1.48

Tapproval Base is the computer half of Tapproval: it puts Claude Code's permission prompts on your Apple Watch, each marked with its risk, and one tap answers. Nothing is approved for you unless you ask it to be, and a CRITICAL never is.

Already installed? Starting Claude Code once updates it.

Since 1.1.47

• A plan on your wrist shows the plan again. Newer Claude Code keeps a plan in a file rather than sending it with the approval request, so the watch could only say "Claude has a plan" with nothing to read. The helper now reads the plan from Claude's own plans folder, so you see its title and steps before you approve.

1.1.47

Tapproval Base is the computer half of Tapproval: it puts Claude Code's permission prompts on your Apple Watch, each marked with its risk, and one tap answers. Nothing is approved for you unless you ask it to be, and a CRITICAL never is.

Already installed? Starting Claude Code once updates it.

Since 1.1.46

• A watch that drops its connection is no longer reported as a crash. When the watch slept or the network blinked mid-request, the helper could show "hit a problem it did not expect" on the Connection screen. Now it notes the hang-up quietly in its log and carries on, while real problems are still reported.

1.1.46

Tapproval Base is the computer half of Tapproval: it puts Claude Code's permission prompts on your Apple Watch, each marked with its risk, and one tap answers. Nothing is approved for you unless you ask it to be, and a CRITICAL never is.

Already installed? Starting Claude Code once updates it.

Since 1.1.45

• Claude's replies read on the wrist the way they do on the phone. Bold stays bold, bullets and numbered lists keep their shape, and sub-points are indented. Seen on the newest watch app; older ones keep the plain reply.

1.1.45

Tapproval Base is the computer half of Tapproval: it puts Claude Code's permission prompts on your Apple Watch, each marked with its risk, and one tap answers. Nothing is approved for you unless you ask it to be, and a CRITICAL never is.

Already installed? Starting Claude Code once updates it.

Since 1.1.44

• Only a real watch can answer. A test build of the watch app running in Apple's Simulator can no longer approve, deny or send words into a live session on your computer. A real watch is unaffected: it is trusted by its own key, as before.

1.1.44

Tapproval Base is the computer half of Tapproval: it puts Claude Code's permission prompts on your Apple Watch, each marked with its risk, and one tap answers. Nothing is approved for you unless you ask it to be, and a CRITICAL never is.

Already installed? Starting Claude Code once updates it.

Since 1.1.43

• A plan reads on the wrist the way it does on the phone. Headings stand out, bold stays bold, and nested steps are indented. Seen on the newest watch app; older ones keep the plain plan.
• A signed-out Mac is said plainly. When Claude Code on your computer is signed out, the watch now shows it as a problem with the one-line fix, not a raw error line.
• A plan never carries a secret to the wrist. Tokens and passwords pasted into a plan are hidden, as they already were everywhere else on a card.

1.1.43

Tapproval Base is the computer half of Tapproval: it puts Claude Code's permission prompts on your Apple Watch, each marked with its risk, and one tap answers. Nothing is approved for you unless you ask it to be, and a CRITICAL never is.

Already installed? Starting Claude Code once updates it.

Since 1.1.42

• A question answered at your computer now leaves the watch. Plans approved at the desk used to stay on the wrist, and so did a question in a long session after Claude Code compacted it. Both now leave within seconds.
• No card waits longer than a day, even when the Mac sleeps in between.

1.1.42

Tapproval Base is the computer half of Tapproval: it puts Claude Code's permission prompts on your Apple Watch, each marked with its risk, and one tap answers. Nothing is approved for you unless you ask it to be, and a CRITICAL never is.

Already installed? Starting Claude Code once updates it.

Since 1.1.41

• The README says what Tapproval does in its first line, and the repository's description and topics now match.
• The website gains llms.txt, a plain summary for AI assistants, and its sitemap dates now follow each page's last change.
• The crash reporter's mail code is shared with the developer's report page. Nothing changes for you.

1.1.41

Tapproval Base is the computer half of Tapproval: it puts Claude Code's permission prompts on your Apple Watch, each marked with its risk, and one tap answers. Nothing is approved for you unless you ask it to be, and a CRITICAL never is.

Already installed? Starting Claude Code once updates it.

Since 1.1.40

• The installer says where to get the watch app. After it finishes, it names the App Store and the TestFlight beta, where the unlock is free.
• The README and the plugin listing mention the beta too.

1.1.40

Tapproval Base is the computer half of Tapproval: it puts Claude Code's permission prompts on your Apple Watch, each marked with its risk, and one tap answers. Nothing is approved for you unless you ask it to be, and a CRITICAL never is.

Already installed? Starting Claude Code once updates it.

Since 1.1.39

• The Connection screen no longer waits on Claude Code's sign-in check. That check can take 15 seconds, and the watch gave up after 5.
• A message or new session sent twice is done once. When the watch gave up waiting and you tapped again, the computer used to do it twice. Now the same words within half a minute get the first answer back.
• New Session lists every project you've worked in, including ones Claude used from a subfolder or a separate worktree.
• The away address recovers by itself when the tunnel starts but never gets an address.
• The computer stays visible on your network: it no longer re-announces itself when nothing changed.
• Steadier reading of usage and activity on a long-running computer.