# Docs for AI agents

> Two kinds of agent work with Trina AX. Coding agents change this repository, and runtime agents answer people inside a team.

AI agents meet Trina AX in two ways. Pick the page for yours.

| | Coding agents | Runtime agents |
|---|---|---|
| **Who** | Claude Code, Codex or another agent changing this repository | Claude Code, Codex or your own Claude, answering people in a team |
| **Works on** | Source code, tests and docs | The team folder, skills and connected services |
| **Bound by** | Repository rules and checks | The asking person's permissions, checked on every call |
| **Start here** | [Working in this repository](/docs/agents/coding-agents.md) | [How AI works inside Trina AX](/docs/agents/runtime.md) |

## For coding agents

- [Working in this repository](/docs/agents/coding-agents.md): Rules, checks, pull request format and what never to commit.
- [Machine-readable docs](/docs/agents/machine-readable-docs.md): `llms.txt`, `llms-full.txt` and a Markdown copy of every page.

## For runtime agents

- [How AI works inside Trina AX](/docs/agents/runtime.md): Engines, tools, permissions, confirmations and limits.
- [Team AX MCP connector](/docs/agents/mcp-connector.md): Connect your own Claude to a team's files over MCP.
- [SKILL.md format](/docs/agents/skills-format.md): The exact header and placement rules for skills.
- [Plugins](/docs/agents/plugins.md): Skill bundles and isolated native tools, installed by admins.