langchain-ai/deepagents · error · PluginManifestError

Could not read plugin manifest {manifest_path}: {exc}

Error message

Could not read plugin manifest {manifest_path}: {exc}

What it means

`load_manifest` reads the manifest with `Path.read_text`; an `OSError` (missing permissions, I/O error, etc. — existence is checked earlier via `_find_manifest_path`) is wrapped as `PluginManifestError` with this message so all manifest failures surface as one exception type.

Source

Thrown at libs/code/deepagents_code/plugins/manifest.py:281

        fallback_name: Name to use only when deriving a manifest-less plugin.

    Returns:
        `(manifest, manifest_path, warnings)`.

    Raises:
        PluginManifestError: If the manifest exists but is invalid.
    """
    manifest_path = find_manifest_path(root)
    if manifest_path is None:
        return None, None, ()
    try:
        decoded = json.loads(manifest_path.read_text(encoding="utf-8"))
    except json.JSONDecodeError as exc:
        msg = f"Invalid JSON syntax in {manifest_path}: {exc}"
        raise PluginManifestError(msg) from exc
    except OSError as exc:
        msg = f"Could not read plugin manifest {manifest_path}: {exc}"
        raise PluginManifestError(msg) from exc
    if not isinstance(decoded, dict):
        msg = f"Plugin manifest {manifest_path} must be a JSON object"
        raise PluginManifestError(msg)
    raw = json_object(decoded)

    warnings: list[str] = []
    name = _validate_name(raw.get("name"), fallback=fallback_name)
    component_paths: dict[str, tuple[Path, ...]] = {}
    for field_name in _PATH_COMPONENT_FIELDS:
        declaration = raw.get(field_name)
        if declaration is None:
            continue
        if field_name in {"mcpServers", "hooks"} and isinstance(declaration, dict):
            continue
        paths = _resolve_component_paths(declaration, root, field_name, warnings)
        if paths:
            component_paths[field_name] = paths

View on GitHub (pinned to a1af029e6e)

Solutions

1. Check permissions on the manifest file shown in the error and make it readable (`chmod u+r`).
2. Verify the file still exists; reinstall the plugin to restore it.
3. Avoid running installs as root so cache files don't become root-owned.
4. Retry if the failure was transient (network filesystem).

Example fix

// before
$ ls -l plugin.json  # -rw------- root root
// after
$ sudo chown $(whoami) plugin.json && chmod u+rw plugin.json

from pathlib import Path
p = Path("plugin.json")
if not p.is_file() or not p.stat().st_mode & 0o444:
    raise SystemExit("manifest missing or unreadable")

try:
    manifest, _, _ = load_manifest(root)
except PluginManifestError as exc:
    if "Could not read plugin manifest" in str(exc):
        print(f"check permissions/restore file: {exc}")
    else:
        raise

When it happens

Trigger: `load_manifest` is called on a plugin root whose manifest file exists at discovery time but cannot be read: permission denied, file deleted between discovery and read, disk I/O error, or path is a broken special file.

Common situations: Manifest file made read-only or owned by another user (e.g. after `sudo` installs); a race where a concurrent update deletes the file; NFS/network mount hiccup; antivirus or backup tool locking the file on Windows.

Related errors

• debug log directory is not owned by the current user: {path}
• Cannot determine whether {str(left)!r} is {str(right)!r}: {e
• Invalid DEEPAGENTS_HOME {str(root)!r}: exists but cannot be
• Failed to read credential file {path}: {exc}. Check the file
• Failed to write credential file {auth_path()}: {exc}. Check

AI-assisted analysis of langchain-ai/deepagents@a1af029e6e (2026-08-29). Data as JSON: /api/errors/8cce64f2433f47b7. Report an issue: GitHub.