cage-flux 0.37.1

Cage — a flux: deterministic attribution ledger for LLM token traffic and tool savings ($0, stdlib).

Cage

Cost dashboards tell you what your AI stack spent. Cage tells you what each tool in it actually saved you — gross and net of the cost of using it — with a method tag on every number.

You're paying for an agent, a graph tool, a rules engine, maybe Copilot. At the end of the month someone asks "is any of this worth it?" — and the honest answer is a shrug and a Slack thread. Cage meters every LLM call, collects a savings receipt from each tool in the stack, and turns the raw stream into an attribution ledger: what you spent, what each tool saved you, what every other combination of tools would have cost, and how much money and time the agent saved versus a person doing the same task. $0, deterministic, zero dependencies, no model in the maintenance path.

Named after John Cage. · Python ≥ 3.11 · stdlib only · MIT · sits beside fux, bach, wagner, orff.

Platforms: macOS is field-validated (real extension sessions, the full manual capture matrix); Linux and Windows are CI-tested across the whole suite + scenario runner. On Windows, run cage doctor --paths first — it shows every log location cage probes on your machine and why any missed. Locked-down endpoint (AppLocker/WDAC blocks the exe, or no pip)? cage setup --python-launcher wires everything through the interpreter instead, and every release ships a single-file cage.pyz.

▶ Demo GIF coming soon.

The story

Another README story. Yeah. Because nobody ever walked out of a meeting humming a feature table, and you will not remember mine. So forget the table. Here's ninety seconds about a conference room, a pile of money, and a bunch of people who have no idea what they're talking about. One of them is you. — Arpit

You ever notice how everybody's saving money now? Everybody. The agent's saving money. The graph tool's saving money. Copilot's saving money. Two tools you built over a weekend — saving money. Add it all up and you should be getting a check in the mail. Funny thing about that. The bill went up.

Here's the con. Nobody — and I mean nobody — can show you the number. They got slides. They got a roadmap. They got a guy named Kevin who "feels like it's a game-changer." What they don't got is one honest figure that says this tool saved this much on this task, and here's what it would've cost to do the boring old way, by hand. Ask for that number and watch the room go quiet and somebody suggest we "circle back."

And the kicker — you built half of it. So when finance points at you and says "is this worth it," you, the expert, the one who's supposed to know — you got a screenshot and a feeling. You're not in trouble for spending the money, folks. You're in trouble because you bought the same fog everybody else did.

Cage is the thing that ruins the fog. It's the itemized receipt nobody asks for and everybody needs: the graph tool saved 27,000 tokens here, fux saved 6,400, the agent did in four minutes what a person does in two hours — plus every other combo you could've run, priced out, each number stamped so you know which ones are real and which ones are some computer's best guess. It doesn't do synergy. It does arithmetic.

See it

$ cage insights matrix --task fix-handover-bug

Counterfactual matrix · task 'fix-handover-bug' · anthropic/claude-opus-4-8
  base 2,000 tok + output 1,500 tok held constant

graphify  fux  compressor   input tok    cost    source
   ✗       ✗       ✗           50,000   $0.1725   modeled
   ✓       ✗       ✗           23,000   $0.0915   modeled
   ✓       ✓       ✗           16,600   $0.0723   modeled
   ✓       ✓       ✓            8,600   $0.0483   measured   ← the run you actually made

  full stack vs all-off: 72% cheaper ($0.1725 → $0.0483)

Per-tool savings any meter can attempt. The part no cost dashboard does is the rest of that table — what each stack you didn't run would have cost — and the source column, so you always know which row is an invoice and which is a reconstruction. Only the configuration you actually ran is measured; no projection ever masquerades as an invoice. That discipline is the whole product.

Quickstart

pip install cage-flux           # the CLI, zero third-party deps
cd your-project
cage setup                      # guided wizard: defaults to all agents, wires skill + hooks + graphify
# non-interactively: cage setup --all   (or --claude / … for just one)
cage demo                       # seed the worked example
cage insights matrix                     # the counterfactual permutation table
cage task quality                    # cost per *successful* task
cage query "how is attribution calculated"  # explain any number — live formula, $0

Adopting into a project — cage setup is the single front door: it offers Claude Code / Copilot / Kiro and defaults to wiring all of them (any agent's hook captures the whole stack, so there's no reason to pick just one). Drive it non-interactively with cage setup --all — or cage setup --claude for a single agent (--no-skill / --no-project / --no-graphify to skip parts). For finer control: cage setup --project-only scaffolds .cage/ + the bin/graphify interceptor without the global skill (agent wiring opt-in via --<agent>), cage setup --wire-only --claude wires just one agent's hooks + MCP, and cage setup --status reports what's already wired.

What gets committed vs what stays local. The project-wired files (.claude/settings.json, .mcp.json, .vscode/mcp.json, .kiro/hooks/) are committed with the repo and contain no absolute paths — they reference the committed shim .cage/bin/cage-run (identical bytes on every machine), which resolves cage at runtime and exits 0 silently when cage isn't installed (a teammate's clone gets working agents, no noise, no capture). Commit .cage/ as-is: its own .gitignore already excludes the machine-local parts (ledger/, out/, state/). Per-machine configs stay absolute and are never cloned: ~/.copilot/hooks/, .git/hooks/ — plus the one committed exception, .kiro/settings/mcp.json (Kiro can't launch MCP servers portably; add it to your .gitignore — cage doctor reminds you). Re-running cage setup migrates any pre-0.20 absolute entries and prints what moved; cage doctor has a portability check.

Metering from your own code is the library adapter — it targets the protocol, not any named client, and is fail-open (a metering error never breaks your call):

import cage

with cage.meter("code-edit", task="fix-bug") as m:
    resp = client.messages.create(...)            # any Anthropic/OpenAI client
    m.usage(provider="anthropic", model="claude-opus-4-8",
            tokens_in=8600, tokens_out=1500, cached_in=3200)

# A tool that shrank the context files a receipt for what it spared you:
cage.record_receipt(tool="fux", raw_alternative=8000, actual=1600,
                    call=m.call_id, task="fix-bug", method="modeled")

Explain it like I'm five

You and a robot helper did the chores. At the end of the day someone wants to know: did the robot actually help, or did it just look busy?

Cage is the chart on the fridge. It writes down how long each chore took with the robot, and how long it would have taken if you'd done it yourself — so you can see, in real minutes and real dollars, which helper earned its place and which one just made noise. And it's careful to mark which numbers it actually timed and which ones are its best guess, so nobody gets fooled by a confident-looking total. It does all of this for free, without ever phoning a friend for the answer.

Why it's different

It's not another cost dashboard. The difference is a set of properties, not features:

• Deterministic. Every derived view — report, attribution, the counterfactual matrix, ROI — is pure parse/arithmetic over an append-only log. Same ledger + same policy ⇒ identical tables, every time. The numbers never drift because nothing guesses.
• Honest by construction. Every figure carries a method: measured (a real invoice), modeled (a reconstructed counterfactual), or estimated (a guess, labelled as one). A projection can never read as an invoice — the one property a "trust me, it paid off" slide can't offer.
• $0 and zero-dependency. Stdlib-only Python, dependencies = []. Heavy ML is an opt-in, off-by-default tier ([embeddings], [ml]), never on the default path. Portable as a tarball, auditable line by line.
• Agent-native. Every read command takes --json; the ledger is served over MCP. Built so an agent can pull its own cost numbers and verify them, not just read a chart.

The "so what" chain: deterministic → so the numbers never hallucinate → so each one carries a defensible method → so you can put the savings claim in front of finance, or an auditor. That last clause is the one a dashboard can't say.

Honest attribution — the part that survives the room

Anyone can sum a bill. Cage's job is to divide credit without lying about it, and it does that with three rules (full design: docs/PLAN.md §4):

• Marginal-by-fixed-order. Each tool's receipt reports the saving it produced given the tools upstream of it; the marginals sum exactly to the total — no overlap, no double-counting, $0 to compute, the order fixed and visible (not a black-box Shapley pass).
• The counterfactual matrix. Cage enumerates the 2ⁿ tool on/off permutations and prices each at the task's model. Only the configuration actually run is measured; every reconstructed cell is modeled (or estimated).
• Quality-adjusted. Cost alone is dishonest — you can "save" by degrading answers and paying for the redo. cage task outcome <task> closes a task with its outcome, and cage task quality reports cost per successful task, the metric that stops false economies.

Counterfactual matrix · task t_9f31 · base model anthropic/claude-sonnet-4-6

compress  graphify  fux    input tok      cost   source
   ✗         ✗       ✗        24,900   $0.0972   modeled
   ✓         ✗       ✗        18,200   $0.0771   modeled
   ✓         ✓       ✗        11,400   $0.0567   modeled
   ✓         ✓       ✓         8,600   $0.0483   measured

full stack vs all-off: ✓ cheaper ($0.0972 → $0.0483)

The savings are anchored to the commit they produced — Cage snapshots a git-aware task record (SHA, branch, diff size, wall-clock) at task close, so a number can always be traced back to the change that earned it.

Authorship — who wrote which commit, and how sure are we

A different question than what did this cost: who is accountable for this diff. cage authorship origin <sha> answers it from the same append-only substrate — which agent wrote which files in which commit, with the same honesty discipline (hooked > transcript > heuristic method ranks; unknown derived from absence, never stored; origin=human only by explicit attestation; CI the sole git-notes writer; counts-never-content — paths and line counts, never a diff body or commit message). Full design: docs/PLAN.md §3.5.

Every number is reviewable — and you can ask it

Cage keeps its numbers in three layers, never mixed, so any figure is auditable in exactly one place:

Layer	Holds	Lives in
Contract	the closed enums (UNITS, METHODS) — the substrate's shape	schema.py
Policy	user-tunable economics: prices, budgets, pipeline order, routing	cage.toml — the only place economic numbers live (previously policy.toml; still read as a fallback)
Constants	code heuristics not meant as config but that must be reviewable: the token divisor, the matrix ceiling, the provenance ranks, the confidence fallback	constants.py

And because the math should explain itself, cage query prints the real formula for any value with its numbers read live from policy + constants — never a hard-coded literal, so an explanation can't drift from the code:

$ cage query "how is attribution calculated"
marginal-attribution · how per-tool savings sum to the total with no double-count
  formula:  walk tools in policy order (graphify → fux → router → compressor → cache
  → response-cache); each receipt is its marginal saving given the tools upstream of
  it, so Σ(marginals) = total, no overlap.
  method:   per-row method = the least-trusted receipt for that tool (honest worst-case).
  code:     cage/attribution.py · cage/matrix.py · cage.toml [tools.order]

Reorder [tools] order in cage.toml and that printed pipeline changes — proof it's the code's actual value, not a slide. It's deterministic and $0: a curated explainer registry, no LLM, no network. cage query also explains how cage itself works (cage query "how does cage work" walks the data flow, attribution, method tags — same live-fact guarantee); cage query --list for every topic, --json for the agent-as-user.

Pricing is managed, and $0 is never silent

A call whose model has no price row bills $0 and says so — report, compare, and study report all print ⚠ N calls (X tokens) UNPRICED — totals understated rather than letting an analyst publish an understated number; the fix is one pasted cage prices set/alias line. Family matching absorbs route prefixes, dotted ids, and effort tiers; prices are derive-time, so fixing the table re-prices every historical row (including imported fleet bundles) retroactively — the ledger stores counts, never conclusions, and cage never fetches a price. The full design — how a call prices · the unpriced workflow · policy versioning and cage prices sync · fleet repricing · the Copilot approximation · credits vs prices — is walked live by cage query prices-cli.

How it works

The same flow as a maintained diagram: docs/architecture-flow.mermaid (renders on GitHub).

One append-only log in, every view derived from it for $0:

record_call / record_receipt  →  .cage/ledger/{calls,receipts,tasks,provenance}.jsonl  (append-only)
        (meter, fail-open)                    │
                                              ▼  derive ($0, no model)
   cage.toml (prices/order/budgets/rates) → report · attrib · matrix · roi
                                             · budget · why · origin

You meter at the provider boundary (library adapter, a reverse proxy for clients you can't edit, or by parsing a Claude Code transcript). Everything downstream is a deterministic projection. The ledger carries token counts, never prompt bodies — PII-safe by construction; point CAGE_LEDGER at a private store to keep even the counts off-disk.

A tool earns rows in attrib/matrix/roi by filing a savings receipt, and there are two ways in, by who owns the tool:

• In-tool (you own it) — e.g. fux carries a fail-open cage_receipt.py and emits its own tool="fux" receipt. Cage stays optional; fux runs unchanged with cage absent.
• External adapter (third-party) — e.g. graphify: cage data graphify -- graphify query "…" runs graphify unmodified, passes its output through byte-for-byte, and files a tool="graphify" receipt by parsing the cited source_files. graphify is never edited; a metering error never alters its result.

The full command surface (30+ subcommands: ledger · attribution · task outcomes · fleet study · ops · agents) is grouped in cage --help, which points at cage query for any "how is this computed". Every read command takes --json for the agent-as-user. The doc map — design of record, subsystem docs, operations, archive — starts at docs/README.md.

Works with any agent — explicit capture over one global ledger

Cage meters whatever speaks the wire format, so all three agents share one ledger contract. Capture is pull-based and universal: cage import reads each agent's on-disk usage log into the ledger, and cage data export refreshes then emits it — they need no hooks, no project, and work the same whether you run a CLI or a VS Code extension.

cage import                 # capture every agent's spend into the active ledger
cage data export --format csv    # refresh, then emit (jsonl | csv | json)
cage report                 # where the spend went
cage data watch                  # optional: a foreground loop you Ctrl-C (no daemon)

The ledger resolves --ledger/CAGE_BASE → project .cage/ → global ~/.cage — so a user with no project captures into the global ledger (cage setup --global to seed it). cage installs no background job (no launchd/systemd/cron); automate it, if you like, with your own cron line calling cage import.

Nonstandard install, a network home, or a custom tool that writes a supported format? Add import paths in cage.toml — [sources.<agent>] paths = ["~/alt/logs"] (or a custom [sources.<name>] format = "claude"), and declare surface = "cli" when a non-IDE store would otherwise be mislabelled; cage doctor --paths shows every location with its provenance and surface, and cage query sources explains the schema.

Agent	Capture (universal)	Optional real-time	Read
Claude Code	cage import (transcript)	Stop hook (CLI only)	cage MCP
Copilot	cage import (session log)	agentStop hook (CLI only)	cage MCP
Kiro	cage import (token log)	agentStop hook (CLI only)	cage MCP
Your code / Orff	cage.meter() library	—	cage CLI / MCP

Hooks are an optional real-time add-on — they fire only under a CLI client, never under a VS Code extension — so cage import/cage data export is the path that always works. cage report --project <name> slices the global ledger by working dir (exact for Claude; Copilot/Kiro logs carry no project, so they're excluded from that filter). Committed wired files never embed a machine's absolute cage path — they reference the repo-local shim .cage/bin/cage-run (see the Quickstart note).

An agent's spend isn't showing up? cage doctor shows the active ledger, each agent's real capture state, and "last import: N ago"; the metadata-only debug log says per agent whether a hook fired or raised (CAGE_DEBUG=1).

Reporting — CSV out of every read view

Every read view also renders as CSV for spreadsheets/BI: --csv streams to stdout (pipe-friendly), --csv <path> writes a file. The same data structure feeds the text table and the CSV, so the numbers can't disagree — and the honesty ships with them: method tags are columns (measured vs estimated survives into the sheet), refusals and the UNPRICED counts stay visible, line endings are LF on every OS (byte-identical, deterministic).

cage report --csv --since 7d > weekly-spend.csv   # last week's spend, flat
cage insights attrib --csv                                  # per-tool savings, method column kept
cage data export --csv calls --since 30d -o calls.csv   # raw ledger rows for a pivot table

--csv works on report · attrib · roi · compare · study report · calibration; raw rows come from cage data export --csv calls|receipts|tasks. CSV is one-way reporting — never an import source; the re-importable fleet bundle stays jsonl (cage data export --study). The cage skill on all four agents knows the recipes — ask your agent for "my weekly cost report as CSV".

The $0 guarantee

Every derived view is parse / arithmetic over the log — no LLM call, ever, on the read or maintenance path. The only model spend is whatever your agent already does; Cage just meters it. The semantic cache and learned compressor ship behind opt-in [embeddings] / [ml] extras; the default install is model-free and dependency-free. 962 tests; cage demo reproduces the worked attribution example against a real ledger.

Honest limits. Marginal-by-fixed-order is defensible and $0, but it is an ordering convention, not a Shapley value (that's a deferred audit mode). And a counterfactual cell is an honest reconstruction, never an invoice — the method column says so on every row, on purpose.

What's new

Latest release below — full history and detail in CHANGELOG.md.

• v0.37.1 (2026-08-01) — Windows dev-CI green. Two same-day follow-ups to v0.36.0: v0.37.0 fixed a release-critical crash (cage setup wrote an unparseable cage.toml on Windows); v0.37.1 fixes the broader dev-CI matrix, which stayed red on windows-latest — cage data graphify crashed on a non-.exe target (subprocess/CreateProcess never honors a #! shebang; now shebang-resolved), plus a few Windows-only test bugs (quote-stripped tokenization, tilde-relative path assertions, a raw backslash in a test's own TOML write). One deeper gap is filed, not fixed: cage's own graphify interceptor is a bash script with no extension, so it's unreachable via a bare graphify PATH lookup on Windows at all — tracked as WIN-GF in docs/OPEN-WORK.md. No substrate/schema/CLI change. See CHANGELOG.md for the full accounting.

The name

Named after John Cage, whose 4′33″ framed four and a half minutes of "silence" so an audience would finally hear the ambient cost they'd been ignoring. Cage the tool does the same to your AI stack: it takes the spend and the savings everyone assumed were free or unknowable, and makes them something you can actually account for. It's part of a family of deterministic substrate → derived views tools — fux (decisions → rules) — and now Cage (LLM traffic + receipts → ledger). The names are deliberate, and they sit beside bach, wagner, and orff.

---

If you've ever been the one in the room with no numbers, run cage demo — pip install cage-flux.

License

MIT — see LICENSE.