Coding agent integrations

In Plain English (For Anyone)	Technical Contract (For Engineers)
What it does: Connects brevio-core directly to Claude Code, Cursor, Codex, Hermes, and OpenClaw so your coding agent checks and repairs its own edits automatically. When to use it: Run brevio-core init or brevio-core hook install in any git repository.	Commands: brevio-core init, brevio-core hook install, brevio-core mcp Catalogue: 123 rules (13 scanned, 110 documented) Exit Codes: The exit code is 0 when nothing is found, 1 when not.

brevio-core checks an agent’s work in three places. Use all three; each catches what the one before let through.

1. In the agent’s loop. A hook scans each file right after the agent edits it, and the uncommitted changes when the agent is about to stop, and hands the findings back to the agent to fix. Where a client has no hook, its rules file and the MCP server do the same job less directly.
2. On the pull request. The Action scans the lines the change adds and fails the check on a finding, whoever or whatever wrote them.
3. In the agent’s instructions. brevio-core init adds a short block of rules to AGENTS.md, which most agents read before they start.

The hook and the MCP server never honour an ignore comment, so an agent cannot silence a finding; only a person can.

Client	Hook after each edit	Skill	MCP server	Instructions
Claude Code	yes	yes	yes	AGENTS.md through CLAUDE.md
Cursor	yes	a rule file	yes	a rule file
Codex	no	no	yes	AGENTS.md
Hermes	no	yes	yes	the skill
OpenClaw	no	yes	yes	the skill

To start a new repository with the check and the rules in place, use the agent-safe starter template (templates/agent-safe-starter).