burnwatch 0.1.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/LICENSE ADDED
@@ -0,0 +1,21 @@
1
+ MIT License
2
+
3
+ Copyright (c) 2026 burnwatch contributors
4
+
5
+ Permission is hereby granted, free of charge, to any person obtaining a copy
6
+ of this software and associated documentation files (the "Software"), to deal
7
+ in the Software without restriction, including without limitation the rights
8
+ to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9
+ copies of the Software, and to permit persons to whom the Software is
10
+ furnished to do so, subject to the following conditions:
11
+
12
+ The above copyright notice and this permission notice shall be included in all
13
+ copies or substantial portions of the Software.
14
+
15
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16
+ IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17
+ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18
+ AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19
+ LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20
+ OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21
+ SOFTWARE.
package/README.md ADDED
@@ -0,0 +1,342 @@
1
+ # burnwatch
2
+
3
+ **Passive cost memory for vibe coding.**
4
+
5
+ burnwatch detects every paid service in your project, tracks what you're spending, and injects budget context directly into your AI coding sessions — so the agent knows what things cost before it recommends burning more money.
6
+
7
+ ```
8
+ ╔══════════════════════════════════════════════════════════════╗
9
+ ║ BURNWATCH — hullscore-app — March 2026 ║
10
+ ╠══════════════════════════════════════════════════════════════╣
11
+ ║ Service Spend Conf Budget Left ║
12
+ ║ ────────────────────────────────────────────────────────── ║
13
+ ║ Anthropic $47.20 ✅ LIVE $100 53% ║
14
+ ║ Vercel $23.00 ✅ LIVE $50 54% ║
15
+ ║ Scrapfly $127.00 ✅ LIVE $50 ⚠️ OVR ║
16
+ ║ Browserbase ~$63.00 🟠 EST $75 16% ║
17
+ ║ Supabase $25.00 ✅ LIVE $100 75% ║
18
+ ║ PostHog ~$49.00 🟡 CALC $49 0% ║
19
+ ╠══════════════════════════════════════════════════════════════╣
20
+ ║ TOTAL: ~$334.20 Untracked: 0 ✅ Est margin: ±$20 ║
21
+ ║ 🚨 SCRAPFLY 254% OVER BUDGET — review before use ║
22
+ ╚══════════════════════════════════════════════════════════════╝
23
+ ```
24
+
25
+ This brief appears automatically at the start of every Claude Code session. You don't open a dashboard. You don't remember to check anything. You just see what you're spending.
26
+
27
+ ---
28
+
29
+ ## Why
30
+
31
+ Agentic development lets you ship 10x faster. It also lets you burn through $400 in Scrapfly credits, rack up unexpected Browserbase bills, and discover PostHog overages three weeks after the code that caused them was written — by an agent, in a session you barely remember.
32
+
33
+ **78% of IT leaders experienced unexpected charges** tied to consumption-based or AI pricing in the past 12 months ([Zylo 2026 SaaS Management Index](https://zylo.com/research/saas-management-index/)).
34
+
35
+ Existing tools either cover one service (ccusage tracks Claude tokens), require enterprise pricing (CloudZero, Vantage), or demand you already know what you're spending on. Nobody watches how services enter your stack and tracks them from the moment of introduction.
36
+
37
+ burnwatch does.
38
+
39
+ ---
40
+
41
+ ## Install
42
+
43
+ ```bash
44
+ # In any project
45
+ npx burnwatch init
46
+ ```
47
+
48
+ That's it. burnwatch scans your project, detects paid services, creates a `.burnwatch/` directory, and registers Claude Code hooks. Next time you start a session, you see your spend.
49
+
50
+ ---
51
+
52
+ ## Quick Start
53
+
54
+ ### 1. Initialize
55
+
56
+ ```bash
57
+ cd your-project
58
+ npx burnwatch init
59
+ ```
60
+
61
+ ```
62
+ 🔍 Scanning project for paid services...
63
+
64
+ Found 11 paid services:
65
+
66
+ • Anthropic (Claude) (✅ LIVE API available)
67
+ Detected via: package.json: @anthropic-ai/sdk, env vars: ANTHROPIC_API_KEY
68
+ • Vercel (✅ LIVE API available)
69
+ Detected via: package.json: @vercel/analytics, @vercel/functions
70
+ • Scrapfly (✅ LIVE API available)
71
+ Detected via: env vars: SCRAPFLY_KEY
72
+ • Supabase (✅ LIVE API available)
73
+ Detected via: package.json: @supabase/supabase-js
74
+ ...
75
+
76
+ 🔗 Registering Claude Code hooks...
77
+ ✅ burnwatch initialized!
78
+ ```
79
+
80
+ ### 2. Add API keys and budgets
81
+
82
+ ```bash
83
+ # LIVE tracking — real billing API data
84
+ burnwatch add anthropic --key $ANTHROPIC_ADMIN_KEY --budget 100
85
+ burnwatch add scrapfly --key $SCRAPFLY_KEY --budget 50
86
+ burnwatch add vercel --token $VERCEL_TOKEN --budget 50
87
+
88
+ # CALC tracking — flat-rate services
89
+ burnwatch add posthog --plan-cost 0 --budget 0
90
+ burnwatch add inngest --plan-cost 25 --budget 25
91
+
92
+ # Just set a budget (tracking stays at detected tier)
93
+ burnwatch add browserbase --budget 75
94
+ ```
95
+
96
+ API keys are stored in `~/.config/burnwatch/` (global, chmod 600). They never touch your project directory. They never end up in git.
97
+
98
+ ### 3. Check your spend
99
+
100
+ ```bash
101
+ burnwatch status
102
+ ```
103
+
104
+ ```
105
+ 📊 Polling services...
106
+
107
+ ╔══════════════════════════════════════════════════════════════╗
108
+ ║ BURNWATCH — hullscore-app — March 2026 ║
109
+ ╠══════════════════════════════════════════════════════════════╣
110
+ ║ Service Spend Conf Budget Left ║
111
+ ║ ────────────────────────────────────────────────────────── ║
112
+ ║ Anthropic $47.20 ✅ LIVE $100 53% ║
113
+ ║ Scrapfly $127.00 ✅ LIVE $50 ⚠️ OVR ║
114
+ ║ Vercel $23.00 ✅ LIVE $50 54% ║
115
+ ║ Supabase $25.00 ✅ LIVE $100 75% ║
116
+ ║ PostHog ~$12.50 🟡 CALC $49 flat — on plan ║
117
+ ║ Browserbase ~$63.00 🟠 EST $75 16% — caution ║
118
+ ╠══════════════════════════════════════════════════════════════╣
119
+ ║ TOTAL: ~$297.70 Untracked: 0 ✅ Est margin: ±$11 ║
120
+ ║ 🚨 SCRAPFLY 254% OVER BUDGET — review before use ║
121
+ ╚══════════════════════════════════════════════════════════════╝
122
+ ```
123
+
124
+ ### 4. Start coding
125
+
126
+ Start a Claude Code session. The spend brief appears automatically. When you mention a tracked service in a prompt, a spend card is injected:
127
+
128
+ ```
129
+ You: "Use Scrapfly to scrape the competitor pricing pages"
130
+
131
+ [BURNWATCH] scrapfly — current period
132
+ Spend: $127.00 | Budget: $50 | ⚠️ 254% over
133
+ Confidence: ✅ LIVE
134
+ ⚠️ 254% of budget consumed
135
+ ```
136
+
137
+ Claude now factors this into its response. It might suggest using Cheerio instead, or warn you about the cost before proceeding.
138
+
139
+ When a new paid service enters your project (new dependency, new env var, new import), burnwatch alerts immediately:
140
+
141
+ ```
142
+ [BURNWATCH] 🆕 New paid service detected: resend
143
+ Run 'burnwatch add resend' to configure budget and tracking.
144
+ ```
145
+
146
+ ---
147
+
148
+ ## How It Works
149
+
150
+ burnwatch runs as Claude Code hooks — background scripts that fire on session events. It never proxies your traffic. It never intercepts API calls. It watches the exhaust of your sessions the same way a court reporter watches a deposition: silently, completely, and without interrupting the work.
151
+
152
+ ### Four Detection Surfaces
153
+
154
+ | Surface | What it catches | When it runs |
155
+ |---------|----------------|-------------|
156
+ | **Package manifest** | `@anthropic-ai/sdk` in dependencies | `init`, `reconcile`, file change |
157
+ | **Environment variables** | `SCRAPFLY_KEY` in process.env | Every session start |
158
+ | **Import statements** | `import { Resend } from "resend"` | `init`, file change |
159
+ | **Prompt mentions** | "use Browserbase to..." | Every prompt |
160
+
161
+ ### Confidence Badges
162
+
163
+ Every spend figure carries an honest confidence badge:
164
+
165
+ | Badge | Meaning | How it works |
166
+ |-------|---------|-------------|
167
+ | ✅ **LIVE** | Real billing API data | Polls service API with your key |
168
+ | 🟡 **CALC** | Fixed monthly cost | You tell burnwatch your plan cost, it projects daily burn |
169
+ | 🟠 **EST** | Instrumented estimate | Usage signals + pricing formula from registry |
170
+ | 🔴 **BLIND** | Detected, not tracked | Service is in your project but no key or cost configured |
171
+
172
+ If burnwatch can't track a service accurately, it says so. The ledger always shows untracked count. You never get a clean dashboard hiding a surprise bill.
173
+
174
+ ### Three Outputs
175
+
176
+ 1. **Session brief** — injected at every session start. Full spend table, alerts, untracked count.
177
+ 2. **Spend cards** — injected when you mention a tracked service. Current spend, budget status, confidence.
178
+ 3. **New service alerts** — injected when a file change introduces a paid service you haven't configured.
179
+
180
+ ### The Ledger
181
+
182
+ burnwatch writes `.burnwatch/spend-ledger.md` at the end of every session — human-readable, git-committable, designed to be read in 10 seconds:
183
+
184
+ ```markdown
185
+ # Burnwatch Ledger — hullscore-app
186
+ Last updated: 2026-03-24T14:32:11Z
187
+
188
+ ## This Month (March 2026)
189
+ | Service | Spend | Conf | Budget | Status |
190
+ |---------|-------|------|--------|--------|
191
+ | Anthropic | $47.20 | ✅ LIVE | $100 | 53% — healthy |
192
+ | Scrapfly | $127.00 | ✅ LIVE | $50 | ⚠️ 254% over |
193
+ | Vercel | $23.00 | ✅ LIVE | $50 | 54% — healthy |
194
+ | PostHog | ~$12.50 | 🟡 CALC | $49 | flat — on plan |
195
+
196
+ ## TOTAL: ~$209.70 (±$2 estimated margin)
197
+ ## Untracked services: 0
198
+ ```
199
+
200
+ ---
201
+
202
+ ## Supported Services (v0.1)
203
+
204
+ | Service | Tier | Billing Model | Notes |
205
+ |---------|------|--------------|-------|
206
+ | Anthropic | ✅ LIVE | Per-token | Requires admin API key |
207
+ | OpenAI | ✅ LIVE | Per-token | Requires org API key |
208
+ | Vercel | ✅ LIVE | Compute + overages | Requires Vercel token |
209
+ | Scrapfly | ✅ LIVE | Credit pool | Standard API key works |
210
+ | Stripe | ✅ LIVE | % of transaction | Tracks processing fees |
211
+ | Supabase | ✅ LIVE | Tiered + overages | Management API |
212
+ | Browserbase | 🟠 EST | Per-session | Estimated from usage |
213
+ | Upstash | 🟠 EST | Per-command | Estimated from usage |
214
+ | Resend | 🟠 EST | Per-email | Estimated from sends |
215
+ | Inngest | 🟡 CALC | Tiered | User-entered plan cost |
216
+ | PostHog | 🟡 CALC | Tiered | User-entered plan cost |
217
+ | Google Gemini | 🟡 CALC | Per-token | User-entered budget |
218
+ | Voyage AI | 🟡 CALC | Per-token | User-entered budget |
219
+ | AWS | 🔴 BLIND | Varies | Detected, complex billing |
220
+
221
+ **Adding a new service?** Edit `registry.json` and open a PR. No release cycle required.
222
+
223
+ ---
224
+
225
+ ## Config Model
226
+
227
+ burnwatch uses a hybrid config model. Sensitive credentials never live in your project directory.
228
+
229
+ ```
230
+ ~/.config/burnwatch/
231
+ config.json ← API keys, tokens (chmod 600, never in git)
232
+
233
+ your-project/.burnwatch/
234
+ config.json ← Tracked services, budgets, detection history
235
+ spend-ledger.md ← Human-readable spend report (git-committable)
236
+ .gitignore ← Ignores cache/snapshots, keeps ledger and config
237
+ data/
238
+ events.jsonl ← Append-only event log
239
+ cache/ ← Billing API response cache
240
+ snapshots/ ← Point-in-time spend snapshots (for delta computation)
241
+ ```
242
+
243
+ ---
244
+
245
+ ## CLI Reference
246
+
247
+ ```bash
248
+ burnwatch init # Initialize in current project
249
+ burnwatch add <service> [options] # Register a service
250
+ burnwatch status # Show current spend brief
251
+ burnwatch services # List all services in registry
252
+ burnwatch reconcile # Scan for untracked services
253
+ burnwatch help # Show help
254
+ burnwatch version # Show version
255
+ ```
256
+
257
+ ### `burnwatch add` options
258
+
259
+ ```bash
260
+ --key <API_KEY> # API key for LIVE tracking
261
+ --token <TOKEN> # Alias for --key
262
+ --budget <AMOUNT> # Monthly budget in USD
263
+ --plan-cost <AMOUNT> # Monthly plan cost (for CALC tracking)
264
+ ```
265
+
266
+ ---
267
+
268
+ ## How the Agent Changes Behavior
269
+
270
+ The real power isn't showing you what you spent — it's telling the agent what everything costs, in context, so cost becomes a factor in every recommendation.
271
+
272
+ When Claude sees `Scrapfly: $127 / $50 budget, 254% over` in its context, it:
273
+
274
+ - Suggests free alternatives (Cheerio, Playwright) before reaching for Scrapfly
275
+ - Warns before generating code that would make more Scrapfly API calls
276
+ - Factors cost into architecture decisions ("this approach would require ~200 more scrape calls")
277
+ - Acknowledges the budget constraint without you having to mention it
278
+
279
+ This feedback loop doesn't exist anywhere else today. The agent has cost memory.
280
+
281
+ ---
282
+
283
+ ## Reconciliation (Sessions Without burnwatch)
284
+
285
+ burnwatch doesn't need to be running in every session. It takes snapshots when it runs and computes deltas between them.
286
+
287
+ ```bash
288
+ burnwatch reconcile
289
+ ```
290
+
291
+ This re-scans your project for services that may have been introduced in sessions where burnwatch wasn't active (via git diffs, new packages, new env vars). Services get flagged and added to tracking.
292
+
293
+ For billing APIs that expose cumulative usage (like Scrapfly's credit counter), burnwatch computes the delta between its last snapshot and the current state — attributing spend across the gap even if it wasn't present for those sessions.
294
+
295
+ ---
296
+
297
+ ## The Registry
298
+
299
+ `registry.json` is the community knowledge base. Each service entry includes:
300
+
301
+ ```json
302
+ {
303
+ "scrapfly": {
304
+ "id": "scrapfly",
305
+ "name": "Scrapfly",
306
+ "packageNames": ["scrapfly-sdk", "scrapfly"],
307
+ "envPatterns": ["SCRAPFLY_KEY", "SCRAPFLY_API_KEY"],
308
+ "importPatterns": ["scrapfly"],
309
+ "mentionKeywords": ["scrapfly"],
310
+ "billingModel": "credit_pool",
311
+ "scalingShape": "linear_burndown",
312
+ "apiTier": "live",
313
+ "pricing": {
314
+ "formula": "credits_used * credit_usd_rate",
315
+ "unitRate": 0.00015,
316
+ "unitName": "credit"
317
+ },
318
+ "gotchas": [
319
+ "Anti-bot bypass options consume 5-25x base credits per request"
320
+ ],
321
+ "alternatives": ["cheerio", "playwright", "firecrawl"],
322
+ "docsUrl": "https://scrapfly.io/docs/scrape-api/billing",
323
+ "lastVerified": "2026-03-24"
324
+ }
325
+ }
326
+ ```
327
+
328
+ The `gotchas`, `alternatives`, and `scalingShape` fields aren't just metadata — the agent reads them and uses them to make better recommendations. Every PR that adds a service makes burnwatch smarter for every user.
329
+
330
+ ---
331
+
332
+ ## Requirements
333
+
334
+ - Node.js 18+
335
+ - Claude Code (for hooks integration)
336
+ - Works without Claude Code too — `burnwatch status` is a standalone CLI
337
+
338
+ ---
339
+
340
+ ## License
341
+
342
+ MIT
package/dist/cli.d.ts ADDED
@@ -0,0 +1 @@
1
+ #!/usr/bin/env node