Step 1 of 3
Your project.
AgentAccord checks your code and your AI agent instructions against rules your team chooses. Start by confirming what it found, then choose rules.
Open a different project
A local folder (including private repositories you have cloned) or a public GitHub repository. Each opens in its own workspace.
Local folder
Use any existing folder, including a private repository you have already cloned. No Git connection is required.
Opening reads the folder. Saving policies or guidance later changes files in this folder.
Public GitHub repository
Download the current default branch as a temporary snapshot. Repository files are never executed.
Contacts GitHub to download source. Public repositories only; no sign-in or token. Changes stay in the snapshot and disappear when this app stops. Limit: 20 MB download, 50 MB expanded, 10,000 entries. Symbolic links are kept but never followed.
Connecting a repository does not mean every language is supported. Review coverage and approve only the rules that fit your project. GitHub snapshots omit history, submodule contents and downloaded Git LFS objects.
Step 2 of 3
Choose your rules.
Nothing is on by default. Turn on only the rules your team wants; your code is never used to guess them.
Share the rules with AI agents
Writes a managed section into AGENTS.md so agents follow the saved rules. Your own text outside that section is kept.
Preview the section
Step 3 of 3
Check.
Checks code and AI agent instructions against the rules you turned on.
Coverage, skipped files, and limitations
Check a SKILL.md or a single file without a project. Your project rules stay unchanged.
1. Choose your input
Text goes only to this local app and temporary files are removed after checking. Up to 20 UTF-8 files, 100 KB each, 400 KB total. Archives are not supported.
No files selected. Paste text below, or select files.
2. Choose rules for this check
Nothing is selected automatically. Markdown supports explicit Java instruction conflicts and Java, Python, and JSON fenced examples. Java files use teaching-example checks. Python uses syntax analysis; JSON checks syntax and duplicate keys. Other languages and application-specific configuration schemas are not yet supported.