.aiignore

.aiignore is a project-level configuration file that tells AI coding agents which files and directories they should leave alone — not read, not send to a model, and not modify. It borrows its pattern-matching convention directly from .gitignore, so a familiar list of globs marks secrets, generated output, or sensitive source as off-limits to the agent. JetBrains adopted it for its Junie coding agent, and the same idea is echoed by parallel files across other AI tools.

• gitignore-style syntax - Uses the same glob patterns developers already know, lowering the cost of adoption.
• Scope control for agents - Keeps credentials, environment files, and proprietary code out of an agent’s working context.
• Repository-local policy - Lives in the project so the rules travel with the code and apply to everyone using the agent.
• A guardrail, not a permission system - It narrows what an agent touches, but it is advisory to the tool rather than an enforced access boundary.

As coding agents move from novelty to daily driver, .aiignore is becoming part of the same hygiene layer as .gitignore and robots.txt — a plain, machine-readable signal that scopes what automation is allowed to see. For API teams, it matters most as a leak-prevention control: keeping .env files, private keys, and internal specs out of the context an agent ships to a model. It sits alongside emerging conventions for governing how agents interact with a codebase, and its presence is a small but telling signal that a team has thought about the agentic turn rather than pointing tools at everything by default.