Xerg is a local-first CLI for finding wasted AI runtime spend. Runtime audits separate three evidence-strict monetary findings from seven neutral signals, report detector coverage, and use to measure compatible changes. Runtime costs may be observed, locally estimated, or unpriced; they are not authoritative provider invoices.

When this file was fetched from rather than loaded from an installed skill:

Always ask before uploading. A local audit needs no account and must not be pushed automatically. After a successful local result, offer to connect and push it with:

npx @xerg/cli@latest activate --push-latest

opens Xerg in the browser. The signed-in user explicitly approves the currently active workspace; the page shows its full organization ID, live plan, and API environment before any credential is issued. If the intended Clerk ID is already known, recommend so both the page and API reject another active workspace. The organization switcher preserves the pairing code. The workspace key is encrypted to the initiating CLI and stored with owner-only, environment-bound metadata; never ask the user to paste a key into chat or expose one in a command.

If approval covers pairing only, use . It must stop after verified credential storage and must not be combined with a source or push flag. A later audit or push requires its own explicit approval.

Codex is an execution host, not a native Xerg audit source. A Codex user audits whatever finds: OpenClaw, Hermes, Claude Code, a Cursor export, or a generic ingest payload.

If the current host is QM, a cold request can explain prerequisites but cannot claim it persisted an admin skill pack, created PostgreSQL privileges, or provisioned secrets. Require an administrator-only private scope. Refuse deployment-wide output in public/shared scopes. Never ask anyone to paste a DSN, identity key, Fly token, provider credential, or Xerg API key into Slack or chat.

Use the non-interactive commands. Do not run unless the user explicitly asks for guided interactive setup — it is TTY-only and prompts for input.

npx @xerg/cli@latest doctor
npx @xerg/cli@latest audit --json

npx @xerg/cli@latest audit --json --compare

If no local data is found, prints the paths it checked. Fallbacks:

If is installed globally, use in place of .

Xerg does not currently ingest provider bills, reconcile invoices, or convert runtime observations into FOCUS. If a user asks for one of those capabilities, explain the boundary and do not present modeled runtime spend as invoice-authoritative.

Costs are priced across input, output, cache read, and cache write tokens using a catalog covering hundreds of current models. Run to see per-file extraction coverage (which economic signals the parser found).

For current Hermes, use ; Xerg prefers . Use for another profile. is mutually exclusive with legacy and . State-only audits retain aggregate spend, request/token/cache, workflow/model/tool, and delegated-workload totals, but first-request cost, initial context, request growth, retry sequences, and identical-input loops are unavailable. The optional plugin writes content-free local events under ; it never adds economics or sends content to Xerg Cloud. Complete observer evidence can split an aggregate only after exact request/token reconciliation.

After installing the observer, restart Hermes, start a new session, and run . Report doctor status and assessed spend. Do not imply that historical aggregate sessions can be reconstructed. Use for CI; unmet coverage exits .

For OpenClaw traces, binds only to , accepts OTLP/HTTP protobuf traces, persists a bounded sanitized capture, and audits after shutdown. It does not modify OpenClaw configuration or push automatically. Without explicit , is independent OpenClaw evidence and cannot be combined with transcript, log, other-runtime, or remote sources.

For certified Hermes traces, uses the same loopback traces-only protocol but HMACs identifiers with a persistent local key and requires . Use only the exact certified commit printed in Xerg's docs. Xerg never installs or edits that plugin. The first-party observer stays primary; cache-inclusive plugin prompt totals are normalized into Xerg's separate input/cache buckets, optional enrichment preserves while analysis identity reflects changed coverage, and conflicting evidence restores the original aggregate. For non-default profiles, use the equivalent environment-variable block printed by the command because the pinned plugin resolves YAML under .

QM currently supports bounded one-shot collection through a host-independent snapshot adapter. An operator outside Slack creates through either a strict view-only direct reader or the Fly-contained one-shot exporter. Fly contained mode uses QM core's existing only inside the hidden exporter, records a process boundary with no database-level least-privilege claim, and never exposes that credential to this skill. Do not mutate setup, install/persist the skill, activate, or push without explicit approval.

In a QM Slack scope, never initiate live database or Fly collection. Accept only an authorized sanitized snapshot already made available in an administrator-only private scope, and use only the exact CLI the operator has explicitly provisioned and version-verified inside that runtime. Current Fly Sprites do not apply the configured sandbox OCI image to persistent Sprites, so never assume an image pin installed the CLI. If it is absent, explain that prerequisite instead of installing it without explicit approval. Continuous follow capture, durable tool-history capture beyond QM's retention window, and live Slack-triggered collection are not currently supported. Never request or receive a DSN, identity key, Fly token, provider credential, or Xerg API key.

Run doctor before every first audit:

xerg doctor --runtime qm --qm-snapshot <snapshot.jsonl>
xerg audit --runtime qm --qm-snapshot <snapshot.jsonl> --since 7d --json

Operator documentation may explain , , , and , but the agent must not execute or provision those live paths from Slack. Direct, Fly, and modes are mutually exclusive.

Explain the result conservatively:

Local snapshot audit is the default. Ask for explicit approval before , , or any other hosted write. Xerg's hosted MCP can read an already-pushed pseudonymous summary but has no native QM database or Fly access.

Local audits need no account. To connect a workspace and push the latest local result, ask permission and run:

npx @xerg/cli@latest activate --push-latest

For a website-first user with no cached audit, securely connects, detects a supported local source, runs the audit, and pushes it. Hosted sync and hosted MCP remain optional and never run without explicit user action.

Use only as a manual recovery path when browser pairing is unavailable. It opens Workspace Settings and masks the key pasted into the terminal. Never ask for or paste a workspace key in agent chat.

Use only for non-interactive CI or deployment automation. Store it in the CI provider's secret manager; never place it inline in a shell command, source file, log, URL, or conversation.