claude-agent-ledger 0.7.0 → 0.7.2

package/README.md

@@ -5,8 +5,9 @@
 [![npm](https://img.shields.io/npm/v/claude-agent-ledger.svg)](https://www.npmjs.com/package/claude-agent-ledger)
 [![license](https://img.shields.io/badge/license-MIT-blue)](LICENSE)
 
-> *Even I needed `--anonymize` before posting my own data publicly.
-> That's why the flag exists. Use it before you screenshot.*
+> Use `--anonymize` before sharing output. It replaces project paths and
+> session IDs, but it does not rename subagent or tool names — review those
+> before you post a screenshot.
 
 ---
 
@@ -26,32 +27,40 @@ If you run **seven** subagents daily (reliability, release, review, iOS factory,
 
 ## Install
 
+Requires [Bun](https://bun.sh/) 1.1 or newer. The CLI is Bun-native even when
+you install it from npm.
+
 ```bash
-npm install -g claude-agent-ledger
-# or: bun install -g claude-agent-ledger
+bun install -g claude-agent-ledger
+# or, with Bun already available: npm install -g claude-agent-ledger
 
 agent-ledger week --summary        # the dashboard (start here)
-agent-ledger week --tree           # subagent fan-out tree (the MOAT view) — new in 0.5
-agent-ledger week --by tool        # which tool ate the budget (Read/Bash/Grep/MCP) — new in 0.5
+agent-ledger week --tree           # parent → child subagent attribution
+agent-ledger week --by tool        # which tool ate the budget (Read/Bash/Grep/MCP)
 agent-ledger week --by model       # opus vs sonnet vs haiku
 agent-ledger week --by project     # which project ate your tokens
 agent-ledger week --by session     # which single sessions went sideways
 agent-ledger week --by day         # daily bar chart with peak/avg/variance
-agent-ledger today --budget 50     # CI gate: exit 2 if today > $50 — new in 0.5
-agent-ledger week --json | jq      # JSON output for ANY view incl. summary — new in 0.5
+agent-ledger today --budget 50     # CI gate: exit 2 if today > $50
+agent-ledger week --json | jq      # JSON output for any view, including summary
 agent-ledger month --md > report.md
 
-# Live tail — polling delta meter — new in 0.7
+# Live tail — polling delta meter
 agent-ledger watch                          # refresh every 10s
 agent-ledger watch 4                        # refresh every 4s
 
-# Root-cause analysis on a single expensive session — new in 0.6
+# Root-cause analysis on a single expensive session
 agent-ledger explain <session-id>           # heuristic: top-3 turns + tool/model breakdown
 ANTHROPIC_API_KEY=sk-ant-... agent-ledger explain <session-id>
-                                            # LLM mode: Haiku 4.5 ≤4-bullet root cause (~$0.001/run)
+                                            # LLM mode: Haiku 4.5 ≤4-bullet root cause (typically <$0.005/run)
 ```
 
-No server. No account. No data leaves your machine.
+All aggregate commands run locally, require no account, and make no network
+requests. `explain` is also local by default. If you explicitly set
+`ANTHROPIC_API_KEY`, it sends derived session metrics — subagent/model/tool
+names, timestamps, token counts, and cost totals — to the
+Anthropic API. The session identifier is redacted. It does **not** send prompts,
+responses, tool content, or project paths.
 
 ## Why this and not `/cost`?
 
@@ -70,77 +79,23 @@ Claude Code's built-in `/cost` shows the **current session** cost. That's it.
 
 Run `agent-ledger week --summary` once and decide.
 
-## Example output (real, from the author's machine)
-
-### `--summary` (the dashboard)
-
-```
-agent-ledger week summary  2026-04-15 → 2026-04-22
-
-  Shadow cost           $13,450.05
-  Sessions              219
-  Projects              9
-  Plan reference        $200/mo (Max)
-  Multiplier            252× vs Max plan
-
-  Top subagent          (main)            $12,787.19 (95%)
-  Top model             claude-opus-4-7   $7,278.87 (54%)
-  Top project           ~/clawbot         $6,657.93 (50%)
-  Peak day              2026-04-15        $3,679.16
-
-  Project mix
-    ~/clawbot                      $6,657.93 (50%)
-    ~/clawbot-work                 $4,433.32 (33%)
-    ~/ZStack-zstack-workspace      $1,890.02 (14%)
-
-  Cache 1h writes       119.65M tokens · $3589.46
-  Cache reads           5.59B tokens · $8236.15
-  Cache reuse           38× reads/writes
-  Server tools          none
-```
-
-### Default — by subagent
-
-```
-agent-ledger week · 2026-04-14 → 2026-04-21
-
-  subagent              sess      tokens(in/out)         cache tokens (1h/5m/read)       cost
-  ───────────────────────────────────────────────────────────────────────────────────────────
-  (main)                  28       514K / 15.87M  1h:116.10M 5m:      0 r:5229.48M  $12524.97
-  general-purpose        148         113K / 955K  1h:      0 5m: 20.67M r: 183.94M    $599.97
-  Swift Developer         13           4K / 574K  1h:      0 5m:  1.61M r:  47.55M     $28.93
-  Market Researcher       10          107K / 74K  1h:      0 5m:   863K r:   8.61M      $7.26
-  Explore                 13          176K / 77K  1h:      0 5m:  2.60M r:  22.38M      $6.47
-  zstackws                 3           232 / 20K  1h:      0 5m:   829K r:   3.52M      $4.47
-  claude-code-guide        1             2K / 2K  1h:      0 5m:    56K r:    223K      $0.08
-  ───────────────────────────────────────────────────────────────────────────────────────────
-  total                  216       916K / 17.57M  1h:116.10M 5m: 26.62M r:5495.71M  $13172.14
-```
+## Example
 
-### `--by model`
+This is deterministic output from the repository's synthetic fixture, using the
+price snapshot documented below:
 
-```
-  subagent              sess      tokens(in/out)         cache tokens (1h/5m/read)       cost
-  ───────────────────────────────────────────────────────────────────────────────────────────
-  claude-opus-4-7         16        59K / 8.35M   1h: 61.15M 5m:  2.32M r:2974.59M   $6967.29
-  claude-opus-4-6        127       544K / 8.25M   1h: 55.02M 5m: 13.30M r:2422.18M   $6160.46
-  claude-sonnet-4-6       65        135K / 946K   1h:      0 5m:  8.81M r:  98.12M     $77.07
-  claude-haiku-4-5        13        178K /  67K   1h:      0 5m:  2.19M r:  21.57M      $4.33
-```
+```text
+agent-ledger today summary
 
-### `--by day` (ASCII bar chart)
+  Shadow cost           $0.08
+  Sessions              2
+  Projects              1
 
+  Top subagent          Reality Checker  $0.06 (74%)
+  Top model             claude-sonnet-4-6  $0.06 (74%)
+  Cache reads           56K tokens · $0.02
+  Cache reuse           28× reads/writes
 ```
-  2026-04-15   $3749.84  ████████████████████████████████████ (18 sess)
-  2026-04-16   $2492.99  ████████████████████████             (175 sess)
-  2026-04-17   $2111.63  ████████████████████                 (9 sess)
-  2026-04-18   $1840.02  ██████████████████                   (18 sess)
-  2026-04-19    $644.97  ██████                               (5 sess)
-  2026-04-20   $1095.58  ███████████                          (6 sess)
-  2026-04-21   $1273.55  ████████████                         (9 sess)
-```
-
-Yes those are real numbers. Yes it's a lot. See [Shadow cost vs actual bill](#shadow-cost-vs-actual-bill).
 
 ## Shadow cost vs actual bill
 
@@ -151,9 +106,13 @@ If you're on a fixed-price subscription (Claude Pro / Max / Claude Code plan), y
 - Understanding where time-and-token budget goes, regardless of how you pay
 - Spotting runaway agents (a 10× spike is still a 10× spike, capped or not)
 - Deciding when to downshift to Haiku or up to Opus per task
-- Justifying a Max subscription internally ("our shadow cost is $13K/wk, Max is $200/mo — that's a 65× discount")
+- Comparing usage concentration with a fixed-price plan. Shadow cost is an
+  estimate, not a claim that subscription usage is interchangeable with raw API
+  capacity.
 
-If you're on pay-as-you-go, the `cost` column is roughly your bill (±5–10% depending on tier / geography / `server_tool_use` calibration).
+If you're on pay-as-you-go, treat the `cost` column as a directional estimate.
+Batch discounts, long-context premiums, provider-specific rates, retries, and
+future price changes can make an invoice differ.
 
 ## How it works
 
@@ -163,10 +122,10 @@ and any `server_tool_use` (web_search / web_fetch) requests. Subagent invocation
 under `<session>/subagents/agent-*.jsonl`, with a sidecar `.meta.json` carrying the real
 subagent name (e.g. `Reality Checker`, `ios-factory`).
 
-`claude-agent-ledger` walks the tree, resolves subagent types from the sidecar metadata,
-and aggregates by your chosen key (subagent, model, or day). Pricing is applied locally
-from `src/pricing.ts`. Nothing is sent anywhere — the tool never touches the network
-after install.
+`claude-agent-ledger` walks the tree, resolves subagent types from the sidecar
+metadata, and aggregates by your chosen key. Pricing is applied locally from
+`src/pricing.ts`. Aggregate commands never touch the network after install; the
+optional API-backed `explain` mode is disclosed in [Install](#install).
 
 ### Cache accounting
 
@@ -182,24 +141,31 @@ Naive trackers lump these together as "input" and either overstate (by counting
 
 ### Server tool pricing
 
-`web_search_requests` and `web_fetch_requests` are counted at $0.01/request and shown in a footer when present:
+`web_search_requests` cost $0.01/request. `web_fetch_requests` are counted for
+observability but have no per-request charge; fetched content still contributes
+normal token costs. Both appear in the footer:
 
 ```
   server tools: web_search ×7, web_fetch ×0  ($0.07)
 ```
 
-If Anthropic's actual rate differs (or web_fetch has its own price), open an issue.
+Rates were checked against Anthropic's official pricing page on 2026-07-23.
+If a log contains a model ID without a verified rate, the CLI warns on stderr
+and leaves that model's token cost at `$0.00` instead of guessing.
 
 ## Flags
 
 | Flag | What it does |
 |---|---|
 | `--summary` | One-screen dashboard with all the headline numbers |
-| `--by <subagent\|model\|day\|project\|session>` | Group rows. Default: `subagent`. `day` renders as ASCII bar chart. |
+| `--by <subagent\|model\|day\|project\|session\|tool>` | Group rows. Default: `subagent`. `day` renders as ASCII bar chart. |
+| `--tree` | Render parent → child subagent cost attribution |
+| `--budget <USD>` | Exit with code 2 when estimated cost exceeds the threshold |
 | `--anonymize` | Replace project paths and session ids with `~/repo-A`, `sess-A`, ... so you can share screenshots without doxing your employer / private repos. |
 | `--md` | Markdown table output (good for committing or sharing) |
 | `--json` | Raw JSON output (pipe into jq) |
-| `--plan pro\|max` | Hide dollar columns; show token utilization only |
+| `--plan pro\|max` | Mask estimated costs in grouped table views; `--summary` and `--tree` keep their cost-oriented output |
+| `--verbose` | Print parse counts and elapsed time to stderr |
 
 | Env | What it does |
 |---|---|
@@ -221,11 +187,16 @@ If Anthropic's actual rate differs (or web_fetch has its own price), open an iss
 
 ## Status
 
-`v0.2.0` shipped 2026-04-21. Validated on ~200K turns of real Claude Code logs.
-Pricing matches Anthropic published rates as of 2026-04-21. Shadow cost values
-may drift ±5-10% from Anthropic's actual invoice in edge cases (retries, tier
-bumps, `server_tool_use` calibration). Calibration PRs welcome — see issue
-templates.
+The npm badge above reports the current published version. See
+[`CHANGELOG.md`](CHANGELOG.md) for release history.
+
+The source price table was checked against Anthropic's published rates on
+2026-07-23. Sonnet 5 automatically switches from its introductory rate to the
+published standard rate on 2026-09-01. Pricing can change, so shadow costs are
+estimates rather than invoice replicas; check
+[`src/pricing.ts`](src/pricing.ts) and the
+[official pricing page](https://platform.claude.com/docs/en/about-claude/pricing)
+before making billing decisions.
 
 - [x] Log parser (JSONL streaming, recursive subagent discovery)
 - [x] Per-subagent aggregation
@@ -235,7 +206,7 @@ templates.
 - [x] ANSI color, TTY-aware
 - [x] Cache 1h/5m TTL split pricing
 - [x] `--plan pro|max` to mask shadow cost
-- [x] `server_tool_use` pricing (web_search, web_fetch)
+- [x] `server_tool_use` accounting (paid web search, no-fee web fetch)
 - [ ] Cost forecasting (trailing 7-day burn rate)
 - [ ] Calibration against real Anthropic invoices
 - [ ] Homebrew tap
@@ -253,14 +224,15 @@ agent-ledger today
 ## Development
 
 ```bash
-bun test           # 22 pass, 0 fail
+bun test           # 35 pass, 0 fail
 bun run typecheck
+bun run build
 ```
 
 ## Author
 
-Built by [@XJM-free](https://jiexiang.dev) while running seven production Claude Code
-subagents for infrastructure work. More context coming on the blog.
+Built by [Jie Xiang](https://jiexiang.dev)
+([@XJM-free](https://github.com/XJM-free)).
 
 ## License
 

package/bin/agent-ledger.ts

@@ -3,8 +3,10 @@ import { Aggregator, type GroupKey, SubagentGraph } from '../src/aggregator.ts';
 import { analyzeSession, formatHeuristic, llmExplain } from '../src/explain.ts';
 import { formatMarkdown, formatSummary, formatTable, formatTrend } from '../src/format.ts';
 import { parseAll } from '../src/parser.ts';
+import { isKnownModel } from '../src/pricing.ts';
 import { formatTree, formatTreeMarkdown } from '../src/tree.ts';
-import type { LedgerReport } from '../src/types.ts';
+import type { LedgerReport, SubagentTreeNode } from '../src/types.ts';
+import packageJson from '../package.json';
 
 type Period = 'today' | 'week' | 'month';
 type Plan = 'payg' | 'pro' | 'max';
@@ -37,8 +39,9 @@ function periodRange(period: Period): { from: Date; to: Date } {
 	return { from, to };
 }
 
-function usage(): never {
-	console.error(`Usage:
+function usage(exitCode = 1): never {
+	const write = exitCode === 0 ? console.log : console.error;
+	write(`Usage:
   agent-ledger <today|week|month> [flags]    — aggregate views
   agent-ledger explain <sessionId>           — root-cause why one session was expensive (new in 0.6)
 
@@ -57,7 +60,7 @@ Flags:
   --budget <USD>              Exit non-zero (code 2) if total cost > USD. Pipe-friendly.
   --md                        Markdown table output (good for sharing / commits)
   --json                      Raw JSON output (also works with --summary now)
-  --plan pro|max              Suppress dollar columns (token utilization only)
+  --plan pro|max              Mask estimated costs in grouped table views.
   --anonymize                 Replace project paths and session ids with safe placeholders.
   --verbose                   Print parse stats (files, turns, elapsed) to stderr.
 
@@ -72,7 +75,7 @@ Examples:
   agent-ledger week --json | jq '.total.cost.totalCost'
   agent-ledger today --budget 50            # CI gate: fail if today > $50
 `);
-	process.exit(1);
+	process.exit(exitCode);
 }
 
 function parseArgs(argv: string[]): Args {
@@ -117,6 +120,16 @@ function parseArgs(argv: string[]): Args {
 			}
 			budget = n;
 			i++;
+		} else if (
+			flag !== '--md' &&
+			flag !== '--json' &&
+			flag !== '--summary' &&
+			flag !== '--tree' &&
+			flag !== '--anonymize' &&
+			flag !== '--verbose'
+		) {
+			console.error(`agent-ledger: unknown flag '${flag}'`);
+			usage();
 		}
 	}
 	return {
@@ -144,6 +157,29 @@ function anonymizeReport(report: LedgerReport, group: GroupKey): LedgerReport {
 	return { ...report, rows: newRows };
 }
 
+function anonymizeTrees(roots: SubagentTreeNode[]): SubagentTreeNode[] {
+	const labels = new Map<string, string>();
+	const labelFor = (sessionId: string): string => {
+		const existing = labels.get(sessionId);
+		if (existing) return existing;
+		const i = labels.size;
+		const suffix =
+			i < 26
+				? String.fromCharCode(65 + i)
+				: String.fromCharCode(65 + Math.floor(i / 26) - 1) +
+					String.fromCharCode(65 + (i % 26));
+		const label = `sess-${suffix}`;
+		labels.set(sessionId, label);
+		return label;
+	};
+	const visit = (node: SubagentTreeNode): SubagentTreeNode => ({
+		...node,
+		sessionId: labelFor(node.sessionId),
+		children: node.children.map(visit),
+	});
+	return roots.map(visit);
+}
+
 function planMask(report: LedgerReport, plan: Plan): LedgerReport {
 	if (plan === 'payg') return report;
 	const zero = () => ({
@@ -231,20 +267,32 @@ async function explainCmd(sessionPrefix: string | undefined): Promise<void> {
 		console.log(llm);
 	} else if (!process.env.ANTHROPIC_API_KEY) {
 		console.log('');
-		console.log('💡 Set ANTHROPIC_API_KEY to enable LLM root-cause analysis (~$0.001/run via Haiku 4.5)');
+		console.log(
+			'💡 Set ANTHROPIC_API_KEY to enable LLM root-cause analysis (typically <$0.005/run via Haiku 4.5)',
+		);
 	}
 }
 
 async function main(): Promise<void> {
+	const cliArgs = process.argv.slice(2);
+	const command = cliArgs[0];
+	if (command === 'help' || cliArgs.includes('--help') || cliArgs.includes('-h')) {
+		usage(0);
+	}
+	if (command === '--version' || command === '-v') {
+		console.log(packageJson.version);
+		return;
+	}
+
 	// Subcommand dispatch.
-	if (process.argv[2] === 'explain') {
+	if (command === 'explain') {
 		return explainCmd(process.argv[3]);
 	}
-	if (process.argv[2] === 'watch') {
+	if (command === 'watch') {
 		const interval = Number(process.argv[3]) || 10;
 		return watchCmd(interval);
 	}
-	const args = parseArgs(process.argv.slice(2));
+	const args = parseArgs(cliArgs);
 	const { from, to } = periodRange(args.period);
 	const t0 = Date.now();
 
@@ -258,8 +306,12 @@ async function main(): Promise<void> {
 	const graph = args.tree ? new SubagentGraph() : undefined;
 
 	let turnCount = 0;
+	const unknownModels = new Set<string>();
 	for await (const turn of parseAll({ from, to })) {
 		turnCount++;
+		if (turn.model && !isKnownModel(turn.model)) {
+			unknownModels.add(turn.model);
+		}
 		aggBySub.add(turn);
 		aggByMod?.add(turn);
 		aggByDay?.add(turn);
@@ -268,6 +320,14 @@ async function main(): Promise<void> {
 		graph?.add(turn);
 	}
 
+	if (unknownModels.size > 0) {
+		console.error(
+			'agent-ledger: no verified price for model(s): ' +
+				[...unknownModels].sort().join(', ') +
+				'; their token cost is shown as $0.00',
+		);
+	}
+
 	if (args.verbose) {
 		const elapsed = Date.now() - t0;
 		console.error(
@@ -278,7 +338,8 @@ async function main(): Promise<void> {
 	let totalCost = aggBySub.finalize().total.cost.totalCost;
 
 	if (args.tree) {
-		const roots = graph!.finalize(20);
+		const rawRoots = graph!.finalize(20);
+		const roots = args.anonymize ? anonymizeTrees(rawRoots) : rawRoots;
 		totalCost = roots.reduce((s, r) => s + r.totalCost, 0);
 		const out =
 			args.json
@@ -308,8 +369,9 @@ async function main(): Promise<void> {
 	} else {
 		let report = aggUser!.finalize();
 		if (args.anonymize) report = anonymizeReport(report, args.by);
-		report = planMask(report, args.plan);
+		// Keep the unmasked estimate for --budget even when --plan hides it.
 		totalCost = report.total.cost.totalCost;
+		report = planMask(report, args.plan);
 		const output =
 			args.json
 				? JSON.stringify(report, null, 2)

package/package.json

@@ -1,15 +1,21 @@
 {
   "name": "claude-agent-ledger",
-  "version": "0.7.0",
+  "version": "0.7.2",
   "description": "See exactly where your Claude Code budget goes — per subagent, per task.",
   "type": "module",
   "bin": {
-    "agent-ledger": "./bin/agent-ledger.ts"
+    "agent-ledger": "bin/agent-ledger.ts"
   },
+  "files": [
+    "bin",
+    "src",
+    "LICENSE",
+    "README.md"
+  ],
   "scripts": {
     "dev": "bun run bin/agent-ledger.ts",
     "test": "bun test",
-    "build": "bun build bin/agent-ledger.ts --target=node --outdir dist",
+    "build": "bun build bin/agent-ledger.ts --target=bun --outdir dist",
     "typecheck": "tsc --noEmit"
   },
   "engines": {
@@ -27,7 +33,11 @@
   "license": "MIT",
   "repository": {
     "type": "git",
-    "url": "https://github.com/XJM-free/claude-agent-ledger.git"
+    "url": "git+https://github.com/XJM-free/claude-agent-ledger.git"
+  },
+  "homepage": "https://github.com/XJM-free/claude-agent-ledger#readme",
+  "bugs": {
+    "url": "https://github.com/XJM-free/claude-agent-ledger/issues"
   },
   "devDependencies": {
     "@types/bun": "latest",

package/src/explain.ts

@@ -5,7 +5,7 @@
 //    tool-call breakdown + cache-reuse stats.
 // 2. LLM mode (set ANTHROPIC_API_KEY): pipes the same evidence into
 //    Haiku 4.5 for a 4-bullet verdict ("prompt bloat", "Opus mis-selection",
-//    "no cache reuse", "tool output overflow"). Costs ~$0.001/run.
+//    "no cache reuse", "tool output overflow"). Typically costs <$0.005/run.
 
 import { homedir } from 'node:os';
 import { join } from 'node:path';
@@ -166,7 +166,9 @@ interface AnthropicMsgRes {
 export async function llmExplain(s: SessionAnalysis): Promise<string | undefined> {
 	const apiKey = process.env.ANTHROPIC_API_KEY;
 	if (!apiKey) return undefined;
-	const evidence = formatHeuristic(s);
+	// The session identifier is useful locally but adds no analytical value to
+	// the API request, so keep it off the network.
+	const evidence = formatHeuristic({ ...s, sessionId: '[redacted]' });
 	const prompt = `You are a Claude Code cost-optimization expert. A user's session burned ${fmtUsd(s.totalCost)} across ${s.turnCount} turns. Below is the per-turn breakdown:
 
 ${evidence}

package/src/pricing.ts

@@ -1,7 +1,7 @@
 import type { CostBreakdown, ModelId, TokenUsage } from './types.ts';
 
 // Prices are per 1M tokens, USD. Keep in sync with
-// https://www.anthropic.com/pricing as of 2026-04-21.
+// https://platform.claude.com/docs/en/about-claude/pricing as of 2026-07-23.
 //
 // Prompt caching (Anthropic):
 //   cache_creation 5-min TTL: 1.25× base input rate
@@ -10,15 +10,31 @@ import type { CostBreakdown, ModelId, TokenUsage } from './types.ts';
 //
 // We include both current and prior-generation model IDs because real logs span
 // version bumps (Claude Code does not auto-rewrite historical logs when you upgrade).
+const SONNET_5_STANDARD_PRICE_START = new Date('2026-09-01T00:00:00Z');
+const SONNET_5_INTRO_RATE = { input: 2, output: 10 };
+const SONNET_5_STANDARD_RATE = { input: 3, output: 15 };
+
 const PRICING: Record<string, { input: number; output: number }> = {
-	// Current generation (2026-04)
-	'claude-opus-4-7': { input: 15, output: 75 },
+	// Current generation (2026-07)
+	'claude-fable-5': { input: 10, output: 50 },
+	'claude-mythos-5': { input: 10, output: 50 },
+	'claude-opus-4-8': { input: 5, output: 25 },
+	'claude-opus-4-7': { input: 5, output: 25 },
+	'claude-opus-4-6': { input: 5, output: 25 },
+	'claude-opus-4-5': { input: 5, output: 25 },
+	// resolveModel applies the introductory rate through 2026-08-31.
+	'claude-sonnet-5': SONNET_5_STANDARD_RATE,
 	'claude-sonnet-4-6': { input: 3, output: 15 },
-	'claude-haiku-4-5-20251001': { input: 0.8, output: 4 },
-	// Prior generation (still appears in logs written before model upgrades)
-	'claude-opus-4-6': { input: 15, output: 75 },
-	'claude-opus-4-5': { input: 15, output: 75 },
 	'claude-sonnet-4-5': { input: 3, output: 15 },
+	'claude-haiku-4-5-20251001': { input: 1, output: 5 },
+	'claude-haiku-4-5': { input: 1, output: 5 },
+	// Prior generation (still appears in logs written before model upgrades)
+	'claude-opus-4-1-20250805': { input: 15, output: 75 },
+	'claude-opus-4-20250514': { input: 15, output: 75 },
+	'claude-sonnet-4-20250514': { input: 3, output: 15 },
+	'claude-haiku-3-5-20241022': { input: 0.8, output: 4 },
+	'claude-3-haiku-20240307': { input: 0.25, output: 1.25 },
+	// Kept for the aliases observed in older Claude Code logs.
 	'claude-haiku-4-0': { input: 0.8, output: 4 },
 };
 
@@ -30,25 +46,30 @@ const CACHE_5M_MULTIPLIER = 1.25;
 const CACHE_1H_MULTIPLIER = 2.0;
 const CACHE_READ_MULTIPLIER = 0.1;
 
-// Anthropic server-tool pricing (https://www.anthropic.com/pricing as of 2026-04):
+// Anthropic server-tool pricing (official pricing docs as of 2026-07-23):
 //   Web search: $10 per 1,000 requests = $0.01 / request
-//   Web fetch:  treated identically here pending Anthropic split (file an issue if wrong)
+//   Web fetch:  no per-request charge; fetched content still incurs token costs
 const WEB_SEARCH_PRICE_PER_REQUEST = 0.01;
-const WEB_FETCH_PRICE_PER_REQUEST = 0.01;
+const WEB_FETCH_PRICE_PER_REQUEST = 0;
 
-function resolveModel(model: ModelId | undefined): { input: number; output: number } {
+function resolveModel(model: ModelId | undefined, asOf: Date): { input: number; output: number } {
 	if (!model) return { input: 0, output: 0 };
 	if (FREE_MODELS.has(model)) return { input: 0, output: 0 };
+	if (model === 'claude-sonnet-5' || model.startsWith('claude-sonnet-5-')) {
+		return asOf < SONNET_5_STANDARD_PRICE_START
+			? SONNET_5_INTRO_RATE
+			: SONNET_5_STANDARD_RATE;
+	}
 	const exact = PRICING[model];
 	if (exact) return exact;
-	// Fuzzy family prefix match as a safety net for unknown subversions.
-	// e.g. "claude-opus-4-8-20260301" falls back to Opus pricing until we add it.
+	// A provider/date-suffixed ID can inherit a verified base model's rate.
+	// Unknown future model families deliberately return zero instead of silently
+	// inventing a price; the CLI warns when it sees one.
 	for (const key of Object.keys(PRICING)) {
-		if (model.startsWith(key)) return PRICING[key] as { input: number; output: number };
+		if (model.startsWith(`${key}-`)) {
+			return PRICING[key] as { input: number; output: number };
+		}
 	}
-	if (model.startsWith('claude-opus-')) return { input: 15, output: 75 };
-	if (model.startsWith('claude-sonnet-')) return { input: 3, output: 15 };
-	if (model.startsWith('claude-haiku-')) return { input: 0.8, output: 4 };
 	return { input: 0, output: 0 };
 }
 
@@ -70,8 +91,12 @@ function splitCacheCreation(usage: TokenUsage): CacheBreakdown {
 	};
 }
 
-export function costFor(usage: TokenUsage, model: ModelId | undefined): CostBreakdown {
-	const rate = resolveModel(model);
+export function costFor(
+	usage: TokenUsage,
+	model: ModelId | undefined,
+	asOf: Date = new Date(),
+): CostBreakdown {
+	const rate = resolveModel(model, asOf);
 	const cacheRead = usage.cache_read_input_tokens ?? 0;
 	const { fiveMin, oneHour } = splitCacheCreation(usage);
 
@@ -110,5 +135,5 @@ export function isKnownModel(model: ModelId | undefined): boolean {
 	if (!model) return false;
 	if (FREE_MODELS.has(model)) return true;
 	if (PRICING[model]) return true;
-	return Object.keys(PRICING).some((key) => model.startsWith(key));
+	return Object.keys(PRICING).some((key) => model.startsWith(`${key}-`));
 }

package/.github/ISSUE_TEMPLATE/bug.md

@@ -1,35 +0,0 @@
----
-name: Bug report
-about: Report incorrect output, crashes, or wrong attribution
-title: ''
-labels: bug
----
-
-## What happened
-
-<!-- What did agent-ledger do? What did you expect it to do instead? -->
-
-## How to reproduce
-
-```bash
-agent-ledger <your-exact-command>
-```
-
-## Output (redact sensitive paths / subagent names)
-
-```
-<paste the output here>
-```
-
-## Your setup
-
-- OS: <!-- macOS 15.3 / Ubuntu 22.04 / ... -->
-- Bun version: <!-- `bun --version` -->
-- agent-ledger version: <!-- `agent-ledger --version` (pre-alpha: use commit SHA) -->
-- Claude Code version: <!-- `claude --version` -->
-
-## If you're on pay-as-you-go, the calibration ask
-
-If you compare `agent-ledger month` against a real Anthropic invoice and
-they differ by > 10%, I want to hear about it. Even one line: "agent-ledger
-says $X, invoice says $Y" is incredibly useful. Thanks.