Compute for AI agents
Giving an AI agent a real machine to work on, inside limits you set.
A coding or automation agent often needs more than an API call: a machine where it can install packages, run tests and keep files between runs. Giving it one raises questions that matter before the first command: how much compute, for how long, with which permissions and at what cost.
This topic collects the tools and pages that help answer them: sizing the compute an agent needs, writing the instructions and permission rules it works under, and keeping its work reviewable.
A recommended path
- Describe the work
List what the agent has to do, which tools it needs and what it must never touch.
- Set the limits
Decide the budget, how long the machine lives and which actions need a person's approval.
- Hand over the context
Give the agent its instructions, files and rules in one place it can read.
Tools and pages
| Page | Type | About |
|---|---|---|
| Agent Permissions Generator | Tool | Choose an agent and a preset and get its permission file with strict defaults: what the agent may do on its own, what it must ask, and what it may never do. Written in your browser, with the tool version and documentation it follows. |
| Agent Infra Risk Assessment | Tool | Answer twelve questions about where your AI agent runs and what it can reach, and get findings by severity, each with a recommendation and the source behind it. Scored in your browser; your answers stay in the link. |
| AGENTS.md generator | Tool | Describe your project and get an AGENTS.md that Codex, Cursor, GitHub Copilot, Claude Code and other coding agents read, with a CLAUDE.md that imports it if you want one. Every section comes from what you write, and nothing leaves your browser. |
| MCP Server Deploy Kit | Tool | Describe your MCP server and get the files to run it on a VPS behind a TLS proxy that streams and checks a bearer token: Compose, Dockerfile, proxy configuration, .env.example and a runbook. Written in your browser; no token is ever written. |
| Agent Handoff Packet | Reading | Hand a tool's files and findings to an AI coding agent as one Markdown task: secrets masked, limits stated, and fixed guardrails against destructive commands and invented secrets. |
See also: AI agents · Methodology · Free tools