diff --git a/Makefile b/Makefile
index b5862e3..ecd1600 100644
--- a/Makefile
+++ b/Makefile
@@ -3,7 +3,7 @@
 # SPDX-License-Identifier: Apache-2.0
 
 SHELL := /bin/bash
-.PHONY: all iso iso-netinst iso-offline package sbom clean test help
+.PHONY: all iso iso-netinst iso-offline package sbom clean test test-search help
 
 # Build configuration
 CODENAME := trixie
@@ -37,6 +37,7 @@ help:
 	@echo "  package PKG=x Build specific package (cx-core, cx-full, cx-archive-keyring)"
 	@echo "  sbom          Generate Software Bill of Materials"
 	@echo "  test          Run build verification tests"
+	@echo "  test-search   Run package search helper tests"
 	@echo "  clean         Remove build artifacts"
 	@echo "  deps          Install build dependencies"
 	@echo ""
@@ -162,6 +163,10 @@ test:
 	./tests/verify-preseed.sh || true
 	@echo -e "$(GREEN)Tests complete$(NC)"
 
+test-search:
+	@echo -e "$(GREEN)Running package search tests...$(NC)"
+	./tests/search-packages-test.sh
+
 # Clean build artifacts
 clean:
 	@echo -e "$(YELLOW)Cleaning build artifacts...$(NC)"
diff --git a/apt/README.md b/apt/README.md
index 3a54e71..c1d3403 100644
--- a/apt/README.md
+++ b/apt/README.md
@@ -81,6 +81,36 @@ git commit -m "Add mypackage 1.0.0"
 git push
 ```
 
+## Smart Package Search
+
+Use `apt/scripts/search-packages.py` to query generated `Packages` indexes before
+publishing or while debugging the repository locally. It supports exact package
+names, short natural-language queries, common synonyms, and small typos without
+requiring network access or extra Python dependencies.
+
+```bash
+# Search generated repository indexes under dists/
+./apt/scripts/search-packages.py "web server"
+
+# Search a specific Packages file
+./apt/scripts/search-packages.py --index dists/stable/main/binary-amd64/Packages postgress
+```
+
+Example output:
+
+```text
+Results:
+  1. postgresql (16+257)
+     object-relational SQL database
+     score=33 reason=fuzzy name; fuzzy term
+```
+
+Run the focused test suite with:
+
+```bash
+make test-search
+```
+
 ### Method 2: Workflow dispatch
 
 Go to Actions → Publish APT Repository → Run workflow
diff --git a/apt/scripts/search-packages.py b/apt/scripts/search-packages.py
new file mode 100755
index 0000000..1341728
--- /dev/null
+++ b/apt/scripts/search-packages.py
@@ -0,0 +1,281 @@
+#!/usr/bin/env python3
+"""Search CX APT package indexes with fuzzy and synonym-aware ranking."""
+
+from __future__ import annotations
+
+import argparse
+import gzip
+import re
+import sys
+from dataclasses import dataclass
+from functools import lru_cache
+from pathlib import Path
+
+
+SYNONYMS = {
+    "database": {"database", "db", "sql", "postgres", "postgresql", "mysql"},
+    "gpu": {"gpu", "nvidia", "amd", "graphics", "cuda", "rocm"},
+    "monitoring": {"monitoring", "metrics", "observability", "prometheus"},
+    "security": {"security", "hardening", "firewall", "sandbox", "secops"},
+    "web": {"web", "http", "server", "nginx", "apache", "caddy"},
+}
+
+SYNONYM_ALIASES = {alias for aliases in SYNONYMS.values() for alias in aliases}
+
+
+@dataclass(frozen=True)
+class Package:
+    name: str
+    version: str
+    description: str
+    fields: dict[str, str]
+
+
+@dataclass(frozen=True)
+class SearchResult:
+    package: Package
+    score: int
+    reason: str
+
+
+@dataclass(frozen=True)
+class QueryContext:
+    normalized: str
+    tokens: set[str]
+    expanded_terms: set[str]
+
+
+def normalize(value: str) -> str:
+    return re.sub(r"[^a-z0-9]+", " ", value.lower()).strip()
+
+
+def tokens(value: str) -> set[str]:
+    return {token for token in normalize(value).split() if token}
+
+
+@lru_cache(maxsize=4096)
+def levenshtein(a: str, b: str) -> int:
+    if a == b:
+        return 0
+    if not a:
+        return len(b)
+    if not b:
+        return len(a)
+
+    previous = list(range(len(b) + 1))
+    for i, char_a in enumerate(a, 1):
+        current = [i]
+        for j, char_b in enumerate(b, 1):
+            current.append(
+                min(
+                    previous[j] + 1,
+                    current[j - 1] + 1,
+                    previous[j - 1] + (char_a != char_b),
+                )
+            )
+        previous = current
+    return previous[-1]
+
+
+def expanded_query_terms(query_terms: set[str]) -> set[str]:
+    expanded = set(query_terms)
+    for term in query_terms:
+        closest_alias = min(SYNONYM_ALIASES, key=lambda alias: levenshtein(term, alias), default="")
+        fuzzy_alias = (
+            closest_alias
+            and len(term) >= 4
+            and term[0] == closest_alias[0]
+            and levenshtein(term, closest_alias) <= 2
+        )
+        if fuzzy_alias:
+            expanded.add(closest_alias)
+
+        for group in SYNONYMS.values():
+            if term in group or (fuzzy_alias and closest_alias in group):
+                expanded.update(group)
+    return expanded
+
+
+def build_query_context(query: str) -> QueryContext:
+    query_tokens = tokens(query)
+    return QueryContext(
+        normalized=normalize(query),
+        tokens=query_tokens,
+        expanded_terms=expanded_query_terms(query_tokens),
+    )
+
+
+def open_index(path: Path):
+    if path.suffix == ".gz":
+        return gzip.open(path, "rt", encoding="utf-8", errors="replace")
+    return path.open("r", encoding="utf-8", errors="replace")
+
+
+def parse_packages_index(path: Path) -> list[Package]:
+    packages: list[Package] = []
+    current: dict[str, str] = {}
+    last_key: str | None = None
+
+    with open_index(path) as handle:
+        for raw_line in handle:
+            line = raw_line.rstrip("\n")
+            if not line:
+                if current:
+                    packages.append(package_from_fields(current))
+                    current = {}
+                    last_key = None
+                continue
+
+            if line.startswith(" ") and last_key:
+                current[last_key] = f"{current[last_key]}\n{line.strip()}"
+                continue
+
+            key, separator, value = line.partition(":")
+            if separator:
+                last_key = key
+                current[key] = value.strip()
+
+    if current:
+        packages.append(package_from_fields(current))
+
+    return packages
+
+
+def package_from_fields(fields: dict[str, str]) -> Package:
+    return Package(
+        name=fields.get("Package", ""),
+        version=fields.get("Version", ""),
+        description=fields.get("Description", ""),
+        fields=dict(fields),
+    )
+
+
+def score_package(package: Package, query: QueryContext) -> SearchResult | None:
+    name_norm = normalize(package.name)
+    package_terms = tokens(f"{package.name} {package.description}")
+
+    score = 0
+    reasons: list[str] = []
+
+    if query.normalized == name_norm:
+        score += 100
+        reasons.append("exact name")
+    elif query.normalized in name_norm:
+        score += 75
+        reasons.append("name contains query")
+
+    overlap = query.expanded_terms & package_terms
+    if overlap:
+        score += 12 * len(overlap)
+        reasons.append("matched " + ", ".join(sorted(overlap)[:4]))
+
+    closest_distance = min(
+        (levenshtein(term, package.name) for term in query.tokens),
+        default=99,
+    )
+    if closest_distance <= 2:
+        score += 45 - (closest_distance * 10)
+        reasons.append("fuzzy name")
+
+    for query_term in query.tokens:
+        if len(query_term) < 4:
+            continue
+        for package_term in package_terms:
+            if len(package_term) < 4:
+                continue
+            if abs(len(query_term) - len(package_term)) > 2:
+                continue
+            if levenshtein(query_term, package_term) <= 2:
+                score += 8
+                reasons.append("fuzzy term")
+                break
+
+    if score <= 0:
+        return None
+
+    return SearchResult(package=package, score=score, reason="; ".join(dict.fromkeys(reasons)))
+
+
+def search(packages: list[Package], query: str, limit: int) -> list[SearchResult]:
+    query_context = build_query_context(query)
+    if not query_context.normalized:
+        return []
+
+    results = [result for package in packages if (result := score_package(package, query_context))]
+    return sorted(results, key=lambda result: (-result.score, result.package.name))[:limit]
+
+
+def default_index_paths(repo_root: Path) -> list[Path]:
+    dists = repo_root / "dists"
+    if not dists.exists():
+        return []
+
+    plain_indexes = sorted(dists.glob("**/Packages"))
+    plain_dirs = {path.parent for path in plain_indexes}
+    gzip_only_indexes = sorted(
+        path for path in dists.glob("**/Packages.gz") if path.parent not in plain_dirs
+    )
+    return plain_indexes + gzip_only_indexes
+
+
+def load_packages(repo_root: Path, indexes: list[Path]) -> list[Package]:
+    paths = indexes or default_index_paths(repo_root)
+    packages: list[Package] = []
+    seen_paths: set[Path] = set()
+    seen_packages: set[tuple[str, str, str]] = set()
+    for path in paths:
+        path_key = path.resolve()
+        if path_key in seen_paths:
+            continue
+        seen_paths.add(path_key)
+
+        if not path.exists():
+            raise FileNotFoundError(f"package index not found: {path}")
+        for package in parse_packages_index(path):
+            package_key = (package.name, package.version, package.description)
+            if package_key in seen_packages:
+                continue
+            seen_packages.add(package_key)
+            packages.append(package)
+    return packages
+
+
+def print_results(results: list[SearchResult]) -> None:
+    if not results:
+        print("No matching packages found.")
+        return
+
+    print("Results:")
+    for index, result in enumerate(results, 1):
+        package = result.package
+        print(f"  {index}. {package.name} ({package.version or 'unknown version'})")
+        if package.description:
+            first_line = package.description.splitlines()[0]
+            print(f"     {first_line}")
+        print(f"     score={result.score} reason={result.reason}")
+
+
+def main(argv: list[str]) -> int:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("query", help="Package name, typo, synonym, or natural-language query")
+    parser.add_argument("--repo-root", type=Path, default=Path.cwd(), help="APT repository root")
+    parser.add_argument("--index", type=Path, action="append", default=[], help="Packages or Packages.gz file")
+    parser.add_argument("--limit", type=int, default=5, help="Maximum results to show")
+    args = parser.parse_args(argv)
+
+    try:
+        packages = load_packages(args.repo_root, args.index)
+    except OSError as error:
+        print(error, file=sys.stderr)
+        return 2
+
+    if not packages:
+        print("No package indexes found. Pass --index or run from an APT repository root.", file=sys.stderr)
+        return 2
+
+    print_results(search(packages, args.query, args.limit))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main(sys.argv[1:]))
diff --git a/tests/search-packages-test.sh b/tests/search-packages-test.sh
new file mode 100755
index 0000000..28cfa09
--- /dev/null
+++ b/tests/search-packages-test.sh
@@ -0,0 +1,114 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+TMP_DIR="$(mktemp -d)"
+trap 'rm -rf "$TMP_DIR"' EXIT
+
+INDEX="$TMP_DIR/Packages"
+DUP_INDEX="$TMP_DIR/Packages.duplicate"
+
+cat > "$INDEX" <<'EOF'
+Package: nginx
+Version: 1.24.0-1
+Description: small, powerful, scalable web server
+
+Package: apache2
+Version: 2.4.58-1
+Description: Apache HTTP server
+
+Package: postgresql
+Version: 16+257
+Description: object-relational SQL database
+
+Package: cx-secops
+Version: 0.1.0-1
+Description: CX Linux security hardening and sandbox tools
+
+Package: cx-gpu-nvidia
+Version: 0.1.0-1
+Description: NVIDIA GPU runtime helpers for CX Linux
+EOF
+gzip -c "$INDEX" > "$INDEX.gz"
+cp "$INDEX" "$DUP_INDEX"
+
+mkdir -p "$TMP_DIR/repo/dists/stable/main/binary-amd64"
+cp "$INDEX" "$TMP_DIR/repo/dists/stable/main/binary-amd64/Packages"
+cp "$INDEX.gz" "$TMP_DIR/repo/dists/stable/main/binary-amd64/Packages.gz"
+
+run_search() {
+    "$ROOT_DIR/apt/scripts/search-packages.py" --index "$INDEX" "$@"
+}
+
+run_repo_search() {
+    "$ROOT_DIR/apt/scripts/search-packages.py" --repo-root "$TMP_DIR/repo" "$@"
+}
+
+assert_contains() {
+    local haystack="$1"
+    local needle="$2"
+
+    if [[ "$haystack" != *"$needle"* ]]; then
+        echo "Expected output to contain: $needle" >&2
+        echo "$haystack" >&2
+        exit 1
+    fi
+}
+
+output="$(run_search postgress)"
+assert_contains "$output" "postgresql"
+assert_contains "$output" "fuzzy"
+
+output="$(run_search "web server")"
+assert_contains "$output" "nginx"
+assert_contains "$output" "apache2"
+
+output="$(run_search "graphics card")"
+assert_contains "$output" "cx-gpu-nvidia"
+
+output="$(run_search "grafics card")"
+assert_contains "$output" "cx-gpu-nvidia"
+
+output="$(run_search hardening)"
+assert_contains "$output" "cx-secops"
+
+output="$("$ROOT_DIR/apt/scripts/search-packages.py" --index "$INDEX.gz" "web server")"
+assert_contains "$output" "nginx"
+
+output="$(run_repo_search postgresql)"
+postgres_count="$(grep -c "^  [0-9][.] postgresql " <<< "$output")"
+if [[ "$postgres_count" -ne 1 ]]; then
+    echo "Expected repo-root search to dedupe Packages and Packages.gz" >&2
+    echo "$output" >&2
+    exit 1
+fi
+
+output="$(run_search "not-a-real-package")"
+assert_contains "$output" "No matching packages found."
+
+output="$(run_search "!!!")"
+assert_contains "$output" "No matching packages found."
+
+output="$("$ROOT_DIR/apt/scripts/search-packages.py" --index "$INDEX" --index "$INDEX" postgresql)"
+postgres_count="$(grep -c "^  [0-9][.] postgresql " <<< "$output")"
+if [[ "$postgres_count" -ne 1 ]]; then
+    echo "Expected duplicate --index paths to be deduped" >&2
+    echo "$output" >&2
+    exit 1
+fi
+
+output="$("$ROOT_DIR/apt/scripts/search-packages.py" --index "$INDEX" --index "$DUP_INDEX" postgresql)"
+postgres_count="$(grep -c "^  [0-9][.] postgresql " <<< "$output")"
+if [[ "$postgres_count" -ne 1 ]]; then
+    echo "Expected overlapping package indexes to be deduped" >&2
+    echo "$output" >&2
+    exit 1
+fi
+
+if "$ROOT_DIR/apt/scripts/search-packages.py" --index "$TMP_DIR/missing" postgresql 2> "$TMP_DIR/missing.err"; then
+    echo "Expected missing index to fail" >&2
+    exit 1
+fi
+assert_contains "$(cat "$TMP_DIR/missing.err")" "package index not found"
+
+echo "search-packages-test.sh: all assertions passed"