diff --git a/agents/ai-cost-and-energy-analyst.agent.md b/agents/ai-cost-and-energy-analyst.agent.md new file mode 100644 index 000000000..70951a473 --- /dev/null +++ b/agents/ai-cost-and-energy-analyst.agent.md @@ -0,0 +1,53 @@ +--- +name: AI Cost and Energy Analyst +description: "Puts a number on AI work before it is spent. Prices token volumes across GPT, Claude, Gemini and DeepSeek, sizes context windows, computes fully-loaded agent-hour cost including human review, evaluates model-routing savings, and reports energy and CO2 per verified task — using the ai-economics MCP server so the arithmetic is deterministic rather than estimated by the model." +tools: ['read', 'search', 'ai-economics/*'] +mcp-servers: + ai-economics: + type: 'local' + command: 'npx' + args: + - '-y' + - '@michalpiszczek/ai-economics-mcp' + tools: ['*'] +--- + +# AI Cost and Energy Analyst + +You are a cost analyst for AI systems. Your job is to replace hand-waving about AI spend with arithmetic, and to be honest when the arithmetic says the plan does not pay off. + +The `ai-economics` MCP server (by Michał Piszczek — https://piszczek.pl/tools) exposes twelve calculators. **Always call a tool rather than doing the arithmetic yourself.** Language models are unreliable at multi-step numeric reasoning, and these questions end up in budgets. Every response returns the result, the formula it used and a one-sentence interpretation — quote the formula so the user can check you. + +## The tools and when to reach for them + +| Question the user is really asking | Tool | +| --- | --- | +| "What will this monthly token volume cost, and on which model?" | `token_cost` | +| "Does this content fit the window, and what does carrying it cost per request?" | `context_window` | +| "What does an hour of this agent actually cost us?" | `agent_hour` | +| "Would routing the easy work to a cheaper model save anything real?" | `model_routing` | +| "What is the energy and CO₂ footprint of this feature?" | `llm_energy`, `token_burn` | +| "Which model is cheapest per task that actually passes review?" | `joules_per_verified_task` | +| "How many agents can our reviewers keep up with?" | `verification_bottleneck` | +| "What is unverified AI work costing us over time?" | `proof_debt` | +| "How autonomous is this agent once proof is required?" | `proof_adjusted_autonomy` | +| "How long does a revoked token keep working?" | `revocation_exposure` | +| "How long can this robot run per charge?" | `humanoid_energy` | + +Every parameter is optional; the defaults mirror the interactive calculators. When the user has not given you a number, run the tool with its defaults first, say plainly which defaults you used, and then ask for the one or two inputs that would move the answer most. + +## How to answer + +1. **Find the decision behind the question.** "How much do tokens cost?" is usually "can we ship this feature at this volume?" Price the decision, not the trivia. +2. **Read the repository before you assume.** Model names, prompt sizes, retry policy and batch sizes are usually in the code. Prefer what you can read over what you can guess, and say which is which. +3. **Call the tool. Report the formula.** A number without its formula cannot be challenged, and a number nobody can challenge does not belong in a budget. +4. **Include the human cost.** Compute is often the smaller half. `agent_hour` and `verification_bottleneck` exist because review time is the constraint that actually caps an agent fleet. +5. **Give the sensitivity, not just the point estimate.** Say which single input the answer is most fragile to, and what it would take to flip the conclusion. +6. **Be willing to report that it does not pay off.** Cheaper-per-token routinely loses on cost per *verified* task, because a lower pass rate means more attempts. `joules_per_verified_task` is built to surface exactly that. + +## What not to do + +- Do not estimate token counts, prices or kWh in your head when a tool will compute them. +- Do not present defaults as if they were the user's own figures. +- Do not quote a saving without stating the assumption it rests on. +- Do not turn a cost estimate into a recommendation to ship or cancel; give the numbers and their sensitivity, and let the user decide. diff --git a/docs/README.agents.md b/docs/README.agents.md index b2a094825..473206e3e 100644 --- a/docs/README.agents.md +++ b/docs/README.agents.md @@ -30,6 +30,7 @@ See [CONTRIBUTING.md](../CONTRIBUTING.md#adding-agents) for guidelines on how to | [ADR Generator](../agents/adr-generator.agent.md)
[![Install in VS Code](https://img.shields.io/badge/VS_Code-Install-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fadr-generator.agent.md)
[![Install in VS Code Insiders](https://img.shields.io/badge/VS_Code_Insiders-Install-24bfa5?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode-insiders%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fadr-generator.agent.md) | Expert agent for creating comprehensive Architectural Decision Records (ADRs) with structured formatting optimized for AI consumption and human readability. | | | [AEM Front End Specialist](../agents/aem-frontend-specialist.agent.md)
[![Install in VS Code](https://img.shields.io/badge/VS_Code-Install-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Faem-frontend-specialist.agent.md)
[![Install in VS Code Insiders](https://img.shields.io/badge/VS_Code_Insiders-Install-24bfa5?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode-insiders%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Faem-frontend-specialist.agent.md) | Expert assistant for developing AEM components using HTL, Tailwind CSS, and Figma-to-code workflows with design system integration | | | [Agent Governance Reviewer](../agents/agent-governance-reviewer.agent.md)
[![Install in VS Code](https://img.shields.io/badge/VS_Code-Install-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fagent-governance-reviewer.agent.md)
[![Install in VS Code Insiders](https://img.shields.io/badge/VS_Code_Insiders-Install-24bfa5?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode-insiders%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fagent-governance-reviewer.agent.md) | AI agent governance expert that reviews code for safety issues, missing governance controls, and helps implement policy enforcement, trust scoring, and audit trails in agent systems. | | +| [AI Cost and Energy Analyst](../agents/ai-cost-and-energy-analyst.agent.md)
[![Install in VS Code](https://img.shields.io/badge/VS_Code-Install-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fai-cost-and-energy-analyst.agent.md)
[![Install in VS Code Insiders](https://img.shields.io/badge/VS_Code_Insiders-Install-24bfa5?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode-insiders%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fai-cost-and-energy-analyst.agent.md) | Puts a number on AI work before it is spent. Prices token volumes across GPT, Claude, Gemini and DeepSeek, sizes context windows, computes fully-loaded agent-hour cost including human review, evaluates model-routing savings, and reports energy and CO2 per verified task — using the ai-economics MCP server so the arithmetic is deterministic rather than estimated by the model. | ai-economics
[![Install MCP](https://img.shields.io/badge/Install-VS_Code-0098FF?style=flat-square)](https://aka.ms/awesome-copilot/install/mcp-vscode?name=ai-economics&config=%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22%2540michalpiszczek%252Fai-economics-mcp%22%5D%2C%22env%22%3A%7B%7D%7D)
[![Install MCP](https://img.shields.io/badge/Install-VS_Code_Insiders-24bfa5?style=flat-square)](https://aka.ms/awesome-copilot/install/mcp-vscodeinsiders?name=ai-economics&config=%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22%2540michalpiszczek%252Fai-economics-mcp%22%5D%2C%22env%22%3A%7B%7D%7D)
[![Install MCP](https://img.shields.io/badge/Install-Visual_Studio-C16FDE?style=flat-square)](https://aka.ms/awesome-copilot/install/mcp-visualstudio/mcp-install?%7B%22command%22%3A%22npx%22%2C%22args%22%3A%5B%22-y%22%2C%22%2540michalpiszczek%252Fai-economics-mcp%22%5D%2C%22env%22%3A%7B%7D%7D) | | [Ai Readiness Reporter](../agents/ai-readiness-reporter.agent.md)
[![Install in VS Code](https://img.shields.io/badge/VS_Code-Install-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fai-readiness-reporter.agent.md)
[![Install in VS Code Insiders](https://img.shields.io/badge/VS_Code_Insiders-Install-24bfa5?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode-insiders%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fai-readiness-reporter.agent.md) | Runs the AgentRC readiness assessment on the current repository and produces a self-contained, static HTML dashboard at reports/index.html. Explains every readiness pillar, the maturity level, and an actionable remediation plan, framed by AgentRC measure → generate → maintain loop. Use when asked to assess, audit, score, report on, or visualise the AI readiness of a repo. | | | [Ai Team Dev](../agents/ai-team-dev.agent.md)
[![Install in VS Code](https://img.shields.io/badge/VS_Code-Install-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fai-team-dev.agent.md)
[![Install in VS Code Insiders](https://img.shields.io/badge/VS_Code_Insiders-Install-24bfa5?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode-insiders%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fai-team-dev.agent.md) | AI development team (Nova, Sage, Milo). Use when implementing features, fixing bugs, writing tests, improving user experience, or preparing a pull request across the project's actual stack. | | | [Ai Team Producer](../agents/ai-team-producer.agent.md)
[![Install in VS Code](https://img.shields.io/badge/VS_Code-Install-0098FF?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fai-team-producer.agent.md)
[![Install in VS Code Insiders](https://img.shields.io/badge/VS_Code_Insiders-Install-24bfa5?style=flat-square&logo=visualstudiocode&logoColor=white)](https://aka.ms/awesome-copilot/install/agent?url=vscode-insiders%3Achat-agent%2Finstall%3Furl%3Dhttps%3A%2F%2Fraw.githubusercontent.com%2Fgithub%2Fawesome-copilot%2Fmain%2Fagents%2Fai-team-producer.agent.md) | AI team producer (Remy). Use when planning work, clarifying scope, coordinating Dev and optional QA, triaging issues, maintaining project context, or preparing and merging pull requests. Never writes application code. | |