Most people who use Claude Code don’t only use Claude Code. You reach for Codex on one task, Amp on another, and half a dozen other agents show up over a month. Each one keeps its own history, in its own format, in its own corner of your disk. To answer a simple question — “how much did I actually spend on AI coding this week?” — you’d have to open five different tools and add it up by hand.

v0.45.0 fixes that. claude-view now reads them all.

One history, whatever ran it

claude-view ingests sessions from 25 other AI coding agents — Amp, Codex, and more — right next to your Claude Code sessions. Same timeline. Same search. Same cost breakdown. You open one app and see everything, regardless of which agent did the work.

That unlocks the questions you actually care about:

Where does my spend go? Cross-provider usage analytics break token and cost totals down per provider, so you can see at a glance whether that expensive run was Claude, Codex, or something else.
Where did I do that thing? Full-text transcript search now spans every provider at once — no more guessing which tool you had open. Find it, then export it.

Cost you can trust — including the newest models

A history is only useful if the numbers are right. Two pieces of this release are about exactly that.

First, sessions on the newest Claude 5-family models — Claude Fable 5 and Claude Sonnet 5 — now show real cost instead of “Unavailable.” Sonnet 5 is priced at its current introductory rate, which means the figure claude-view shows you is the figure you’re actually billed today, not a list price you’re not paying. When the introductory period ends, the number updates with it. 寧願唔顯示，都唔顯示錯嘅嘢 — better to show nothing than something wrong, and better still to show the true number.

Second, if you spin up agents in the background, their goals now ride along on the live card. A running session shows what it’s working toward without you having to open it.

Under the hood

Reading 25 formats without turning claude-view into a pile of special cases meant normalizing every provider into the same internal session shape the Claude Code pipeline already uses — so cost, search, and the timeline all work identically no matter the source. Claude Code’s own path is untouched; the other providers plug into the same rails.

We also taught the parser about Claude Code’s newer bridge and background-session records, so a CLI update can’t quietly slip a new field past the schema guard — and added a session check that flags exposure to the active Shai-Hulud / Miasma / Hades npm worm campaign going after AI-tool developers.

What’s next

Deeper per-provider insights — trends and comparisons, not just totals.
More agents as their formats stabilize.
Richer live monitoring for background and bridged sessions.

Update now

npx claude-view@latest

Update, then open claude-view — if you’ve used any agent besides Claude Code, its sessions are already there waiting for you.