# Frazer for agents

> How an AI agent reads Frazer's docs and uses Frazer, the frazer CLI and Fir correctly: the Markdown files and indexes, the setup steps, the rules every write follows, the corrections to what models assume, and procedures with their commands, their checks and the step that needs a person.

Part of Frazer's docs for agents. Every page on frazer.ai: https://frazer.ai/llms.txt. Every docs page: https://frazer.ai/docs/llms.txt.

## How agents should use Frazer for agents

- Fetch the Markdown version of a page (its path plus `.md`), and only the index and pages your task needs.
- Plan every change first with `--dry-run`, and ask before you change a person's resources: writes through MCP wait for a person's approval, and agents never approve.
- Open a Fir session before you change code (`fir session open`, or your harness's Fir plugin); an agent's commit without one is refused with `SESSION_REQUIRED`.
- Never paste a secret into a prompt, a file or a commit: request a person task with a secret slot, and receive a reference.

## Pages

- [Frazer for agents](https://frazer.ai/docs/for-agents.md?from=llms-txt): How an AI agent reads Frazer's docs and uses Frazer, the frazer CLI and Fir correctly: the Markdown files, the setup steps, the rules every write follows and the facts models most often get wrong.
- [Tasks, step by step](https://frazer.ai/docs/for-agents/tasks.md?from=llms-txt): The tasks agents are asked for most on Frazer: create a project, deploy, add a secret, open a Fir session, open and merge a pull request on Fir, draft a plan and improve it in a loop, work on its items, ask a person for what only a person can do, build a stack for a plan, land your own fix at hotfix priority, resolve a conflicted layer, roll back, add a server, each as numbered commands with a check and the step that needs a person.

## Optional

- [Full text of these pages](https://frazer.ai/docs/for-agents/llms-full.txt): 2 pages, 37,242 characters, about 9,311 tokens