diff --git a/Cargo.toml b/Cargo.toml
index d4e6f04..2c4604f 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -29,6 +29,10 @@ tempfile = "3.10"
 name = "staging_bench"
 harness = false
 
+[[bench]]
+name = "app_bench"
+harness = false
+
 # cargo-dist configuration
 [workspace.metadata.dist]
 # CI backends
diff --git a/benches/app_bench.rs b/benches/app_bench.rs
new file mode 100644
index 0000000..e98dcfa
--- /dev/null
+++ b/benches/app_bench.rs
@@ -0,0 +1,181 @@
+//! Benchmarks for git-twig application operations
+//!
+//! These benchmarks measure the performance of core operations:
+//! - Tree flattening (Node::flatten)
+//! - Filter nodes (App::filter_nodes)
+//! - ANSI code stripping (strip_ansi_codes)
+//! - Full refresh cycle
+
+use criterion::{black_box, criterion_group, criterion_main, BenchmarkId, Criterion};
+use std::collections::HashSet;
+
+use git_twig::node::{Node, NodeType};
+use git_twig::theme::{Theme, ThemeType};
+use git_twig::tui::app::{strip_ansi_codes, App};
+
+/// Create a test tree with the specified depth and breadth
+fn create_test_tree(depth: usize, breadth: usize) -> Node {
+    fn build_level(current_depth: usize, max_depth: usize, breadth: usize) -> Node {
+        if current_depth >= max_depth {
+            // Create a file node using the constructor
+            Node::new_file(
+                format!("file_{}.rs", current_depth),
+                format!("path/to/file_{}.rs", current_depth),
+                "M".to_string(),
+                Some((10, 5)), // (added, deleted)
+            )
+        } else {
+            // Create a directory with children
+            let mut children = Vec::new();
+            for _ in 0..breadth {
+                let child = build_level(current_depth + 1, max_depth, breadth);
+                children.push(child);
+            }
+            // Add some file nodes at each level
+            for i in 0..3 {
+                children.push(Node::new_file(
+                    format!("file_{}.rs", i),
+                    format!("level_{}/file_{}.rs", current_depth, i),
+                    "M".to_string(),
+                    Some((i + 1, i)),
+                ));
+            }
+            Node::new_dir(
+                format!("dir_{}", current_depth),
+                format!("path/to/dir_{}", current_depth),
+                children,
+            )
+        }
+    }
+
+    build_level(0, depth, breadth)
+}
+
+/// Benchmark tree flattening with various tree sizes
+fn bench_tree_flatten(c: &mut Criterion) {
+    let mut group = c.benchmark_group("tree_flatten");
+    let theme = Theme::new(ThemeType::Unicode);
+    let collapsed_paths: HashSet<String> = HashSet::new();
+
+    // Test with different tree configurations (depth, breadth)
+    for (depth, breadth) in [(3, 3), (4, 4), (5, 3), (3, 10)].iter() {
+        let tree = create_test_tree(*depth, *breadth);
+        let node_count = count_nodes(&tree);
+
+        group.bench_with_input(BenchmarkId::new("nodes", node_count), &tree, |b, tree| {
+            b.iter(|| {
+                black_box(tree.flatten(2, false, &theme, &collapsed_paths));
+            });
+        });
+    }
+
+    group.finish();
+}
+
+/// Count total nodes in a tree
+fn count_nodes(node: &Node) -> usize {
+    match &node.node_type {
+        NodeType::File { .. } => 1,
+        NodeType::Directory { children } => 1 + children.iter().map(count_nodes).sum::<usize>(),
+    }
+}
+
+/// Benchmark filter_nodes with various query lengths and node counts
+fn bench_filter_nodes(c: &mut Criterion) {
+    let mut group = c.benchmark_group("filter_nodes");
+    let theme = Theme::new(ThemeType::Unicode);
+    let collapsed_paths: HashSet<String> = HashSet::new();
+
+    // Create trees of different sizes
+    for (depth, breadth) in [(3, 5), (4, 4), (5, 3)].iter() {
+        let tree = create_test_tree(*depth, *breadth);
+        let flat_nodes = tree.flatten(2, false, &theme, &collapsed_paths);
+        let node_count = flat_nodes.len();
+
+        // Test with different query lengths
+        for query in ["", "file", "dir_2", "nonexistent"].iter() {
+            let id = format!("{}_nodes_query_{}", node_count, query.len());
+            group.bench_with_input(BenchmarkId::new("filter", &id), &flat_nodes, |b, nodes| {
+                b.iter(|| {
+                    black_box(App::filter_nodes(nodes, query));
+                });
+            });
+        }
+    }
+
+    group.finish();
+}
+
+/// Benchmark ANSI code stripping
+fn bench_strip_ansi_codes(c: &mut Criterion) {
+    let mut group = c.benchmark_group("strip_ansi_codes");
+
+    // Sample strings with ANSI codes (simulating diff output)
+    let test_strings: Vec<String> = vec![
+        // No ANSI codes
+        "plain text without any formatting".to_string(),
+        // Single ANSI code
+        "\x1B[32m+added line\x1B[0m".to_string(),
+        // Multiple ANSI codes
+        "\x1B[31m-\x1B[0m\x1B[31mdeleted line with \x1B[1mbold\x1B[0m text\x1B[0m".to_string(),
+        // Long line with many codes
+        (0..50)
+            .map(|i| {
+                if i % 2 == 0 {
+                    format!("\x1B[32m+line {}\x1B[0m", i)
+                } else {
+                    format!("\x1B[31m-line {}\x1B[0m", i)
+                }
+            })
+            .collect::<Vec<_>>()
+            .join("\n"),
+    ];
+
+    for (idx, s) in test_strings.iter().enumerate() {
+        let id = format!("len_{}_idx_{}", s.len(), idx);
+        group.bench_with_input(BenchmarkId::new("strip", &id), s, |b, s| {
+            b.iter(|| {
+                black_box(strip_ansi_codes(s));
+            });
+        });
+    }
+
+    group.finish();
+}
+
+/// Benchmark multiple strip_ansi_codes calls (simulating diff search)
+fn bench_strip_ansi_batch(c: &mut Criterion) {
+    let mut group = c.benchmark_group("strip_ansi_batch");
+
+    // Simulate a diff with many lines
+    let diff_lines: Vec<String> = (0..500)
+        .map(|i| {
+            if i % 3 == 0 {
+                format!("\x1B[32m+added line {} with some content\x1B[0m", i)
+            } else if i % 3 == 1 {
+                format!("\x1B[31m-deleted line {} with content\x1B[0m", i)
+            } else {
+                format!(" context line {} unchanged", i)
+            }
+        })
+        .collect();
+
+    group.bench_function("500_lines", |b| {
+        b.iter(|| {
+            for line in &diff_lines {
+                black_box(strip_ansi_codes(line));
+            }
+        });
+    });
+
+    group.finish();
+}
+
+criterion_group!(
+    benches,
+    bench_tree_flatten,
+    bench_filter_nodes,
+    bench_strip_ansi_codes,
+    bench_strip_ansi_batch,
+);
+criterion_main!(benches);
diff --git a/src/lib.rs b/src/lib.rs
new file mode 100644
index 0000000..f9c640e
--- /dev/null
+++ b/src/lib.rs
@@ -0,0 +1,12 @@
+//! git-twig library for benchmarking and testing
+//!
+//! This module exposes internal types for benchmarking purposes.
+
+pub mod cache;
+pub mod config;
+pub mod git;
+pub mod icons;
+pub mod node;
+pub mod parser;
+pub mod theme;
+pub mod tui;
diff --git a/src/tui/app.rs b/src/tui/app.rs
index 999ae45..c0fa888 100644
--- a/src/tui/app.rs
+++ b/src/tui/app.rs
@@ -1199,7 +1199,8 @@ impl App {
 static ANSI_RE: LazyLock<regex::Regex> =
     LazyLock::new(|| regex::Regex::new(r"\x1B\[[0-9;]*[mK]").unwrap());
 
-fn strip_ansi_codes(s: &str) -> String {
+/// Strip ANSI escape codes from a string
+pub fn strip_ansi_codes(s: &str) -> String {
     ANSI_RE.replace_all(s, "").to_string()
 }
 
diff --git a/src/tui/mod.rs b/src/tui/mod.rs
index 85d9dad..7d3963b 100644
--- a/src/tui/mod.rs
+++ b/src/tui/mod.rs
@@ -6,7 +6,7 @@ use crossterm::{
 use ratatui::{backend::CrosstermBackend, Terminal};
 use std::io;
 
-mod app;
+pub mod app;
 mod event;
 mod history;
 mod ui;