# Xerg

> Audit AI agent runtime spend in dollars. Runtime audits report identified waste plus assessed detector spend, so aggregate-only data is never presented as a conclusive zero. Costs may be observed, locally estimated, or unpriced; they are not provider invoices. Start local with the CLI, save the last 30 days of pushed audits in a free workspace, and upgrade to Pro for full history, Ask Xerg, hosted MCP, Slack, and policies.

## Docs

- [Documentation home](https://xerg.ai/docs): Xerg docs index
- [Quickstart](https://xerg.ai/docs/quickstart): Run the init-first local flow, compare fixes, and add hosted follow-up only if needed
- [Installation](https://xerg.ai/docs/installation): Install and run your first audit
- [Doctor](https://xerg.ai/docs/doctor): Verify your local or remote setup
- [Audit](https://xerg.ai/docs/audit): Run local audits and reports
- [FOCUS retirement](https://xerg.ai/docs/focus): Understand the retired importer and the runtime-versus-invoice boundary
- [Hermes analysis](https://xerg.ai/docs/hermes-analysis): Read state.db, reconcile first-party observer requests, and inspect explicit coverage
- [Hermes trace collection](https://xerg.ai/docs/hermes-traces): Receive certified sanitized local trace enrichment with state.db authority
- [OpenClaw trace collection](https://xerg.ai/docs/openclaw-traces): Receive a bounded sanitized local trace capture without automatic push
- [QM secure setup](https://xerg.ai/docs/qm-secure-setup): Choose offline snapshot, strict direct-reader, or Fly-contained collection
- [QM collection](https://xerg.ai/docs/qm-collection): Collect and audit bounded one-shot QM snapshots
- [QM pricing and limitations](https://xerg.ai/docs/qm-pricing-coverage): Understand explicit-model pricing, openrouter/auto, and coverage gates
- [Hosted dashboard](https://xerg.ai/docs/hosted-dashboard): Use pushed audits, sources, trends, Optimizations, policies, and API keys
- [Auth and push](https://xerg.ai/docs/auth-and-push): Push audits to a hosted workspace (Free or Pro)
- [Hosted MCP](https://xerg.ai/docs/hosted-mcp): Connect remote MCP clients to hosted workspace data
- [Optimizations and policies](https://xerg.ai/docs/optimizations-and-policies): Track recommendation-backed work and policy drafts
- [Agent clients](https://xerg.ai/docs/agent-clients): Configure QM, Codex, Cursor, Claude Code, Claude Desktop, and generic MCP clients
- [Service status](https://status.xerg.ai): Current Dashboard, API, Hosted MCP, Slack Integration, Website, and Documentation availability

## Links

- [For Agents](https://xerg.ai/agents): Agent setup hub
- [AI Agent Spend Scenario Modeler](https://xerg.ai/agent-spend-modeler): Explore an adjustable, illustrative allocation of current agent spend into useful work and modeled waste categories; it is not an audit finding or savings promise
- [MCP server](https://xerg.ai/mcp): Hosted MCP endpoint for Pro and Enterprise workspaces
- [SKILL.md](https://xerg.ai/skill.md): Agent skill definition for Claude Code, Cursor, and other agent ecosystems
- [Full docs for LLMs](https://xerg.ai/llms-full.txt): Complete reference in a single document
- [npm: @xerg/cli](https://www.npmjs.com/package/@xerg/cli): Public CLI package, not open source
- [npm: @xerg/schemas](https://www.npmjs.com/package/@xerg/schemas): MIT-licensed public wire-contract types
- [Dashboard](https://xerg.ai/dashboard): Hosted dashboard (free account required)
- [Pricing](https://xerg.ai/pricing): Free, Pro, and Enterprise plans
- [Status](https://status.xerg.ai): Xerg Cloud and public web availability and incident history