.biggiignore

The primary mechanism for controlling file access is the permission system in biggi.jsonc. You define tool-level permissions with glob patterns:

{
  "permission": {
    "read": { "*.env": "deny", "*": "allow" },
    "edit": { "dist/**": "deny", "*": "allow" }
  }
}

If you have an existing .biggiignore file, it is still supported. The IgnoreMigrator automatically converts .biggiignore patterns into permission deny rules on read and edit tools, so your existing rules continue to work without manual changes.

You can also exclude paths from the file watcher separately using watcher.ignore:

{
  "watcher": {
    "ignore": ["tmp/**", "logs/**"]
  }
}

File access is controlled through permission-based access control. Each tool (read, edit, glob, grep, write, bash, etc.) has its own permission rules evaluated against glob patterns.

In addition to your explicit permission rules:

• Hardcoded directory ignores — 27 directories are always skipped (e.g. node_modules, .git, dist, build, .cache, __pycache__, vendor, and others).
• Hardcoded file pattern ignores — 11 file patterns are always skipped (e.g. lock files, binary artifacts).
• .gitignore and .ignore files are also respected when listing and searching files.

If a file is denied by a permission rule, the tool will report that access was blocked.

Permission Rules

Permission rules are defined per-tool in biggi.jsonc. Patterns are evaluated in order — the last matching rule wins:

{
  "permission": {
    "read": {
      "*.env": "deny",
      "secrets/**": "deny",
      "*": "allow"
    },
    "edit": {
      "dist/**": "deny",
      "*.lock": "deny",
      "*": "allow"
    }
  }
}

Migrating from .biggiignore

If you already have a .biggiignore file, you don't need to do anything — the IgnoreMigrator reads your existing patterns and applies them as deny rules on read and edit tools automatically. You can optionally move your rules into biggi.jsonc for more granular control (e.g. denying edits but allowing reads).

File Watcher Exclusions

The watcher.ignore setting controls which paths the file watcher skips. This is separate from tool permissions and only affects change detection:

{
  "watcher": {
    "ignore": ["tmp/**", "logs/**", ".build/**"]
  }
}