diff --git a/air/Cargo.toml b/air/Cargo.toml
index e04cecbef1..7db8686ff5 100644
--- a/air/Cargo.toml
+++ b/air/Cargo.toml
@@ -27,6 +27,7 @@ harness = false
 [features]
 default = ["std"]
 std = ["vm-core/std", "winter-air/std"]
+internals = []
 
 [dependencies]
 vm-core = { package = "miden-core", path = "../core", version = "0.8", default-features = false }
diff --git a/air/src/trace/main_trace.rs b/air/src/trace/main_trace.rs
index 56157ea350..699e38bf5d 100644
--- a/air/src/trace/main_trace.rs
+++ b/air/src/trace/main_trace.rs
@@ -16,7 +16,9 @@ use super::{
     CHIPLETS_OFFSET, CLK_COL_IDX, CTX_COL_IDX, DECODER_TRACE_OFFSET, FMP_COL_IDX, FN_HASH_OFFSET,
     STACK_TRACE_OFFSET,
 };
-use core::ops::Range;
+use core::ops::{Deref, Range};
+#[cfg(any(test, feature = "internals"))]
+use vm_core::utils::collections::Vec;
 use vm_core::{utils::range, Felt, ONE, ZERO};
 
 // CONSTANTS
@@ -28,12 +30,20 @@ const DECODER_HASHER_RANGE: Range<usize> =
 // HELPER STRUCT AND METHODS
 // ================================================================================================
 
-pub struct MainTrace<'a> {
-    columns: &'a ColMatrix<Felt>,
+pub struct MainTrace {
+    columns: ColMatrix<Felt>,
 }
 
-impl<'a> MainTrace<'a> {
-    pub fn new(main_trace: &'a ColMatrix<Felt>) -> Self {
+impl Deref for MainTrace {
+    type Target = ColMatrix<Felt>;
+
+    fn deref(&self) -> &Self::Target {
+        &self.columns
+    }
+}
+
+impl MainTrace {
+    pub fn new(main_trace: ColMatrix<Felt>) -> Self {
         Self {
             columns: main_trace,
         }
@@ -43,6 +53,14 @@ impl<'a> MainTrace<'a> {
         self.columns.num_rows()
     }
 
+    #[cfg(any(test, feature = "internals"))]
+    pub fn get_column_range(&self, range: Range<usize>) -> Vec<Vec<Felt>> {
+        range.fold(vec![], |mut acc, col_idx| {
+            acc.push(self.get_column(col_idx).to_vec());
+            acc
+        })
+    }
+
     // SYSTEM COLUMNS
     // --------------------------------------------------------------------------------------------
 
diff --git a/processor/Cargo.toml b/processor/Cargo.toml
index 681f2b544a..caae5e15a1 100644
--- a/processor/Cargo.toml
+++ b/processor/Cargo.toml
@@ -19,7 +19,7 @@ doctest = false
 [features]
 concurrent = ["std", "winter-prover/concurrent"]
 default = ["std"]
-internals = []
+internals = ["miden-air/internals"]
 std = ["tracing/attributes", "vm-core/std", "winter-prover/std"]
 
 [dependencies]
diff --git a/processor/src/chiplets/aux_trace/mod.rs b/processor/src/chiplets/aux_trace/mod.rs
index 85bacfddda..2aceaff772 100644
--- a/processor/src/chiplets/aux_trace/mod.rs
+++ b/processor/src/chiplets/aux_trace/mod.rs
@@ -1,4 +1,4 @@
-use super::{super::trace::AuxColumnBuilder, ColMatrix, Felt, FieldElement, StarkField, Vec};
+use super::{super::trace::AuxColumnBuilder, Felt, FieldElement, StarkField, Vec};
 
 use miden_air::trace::{
     chiplets::{
@@ -56,7 +56,7 @@ impl AuxTraceBuilder {
     /// provided by chiplets in the Chiplets module.
     pub fn build_aux_columns<E: FieldElement<BaseField = Felt>>(
         &self,
-        main_trace: &ColMatrix<Felt>,
+        main_trace: &MainTrace,
         rand_elements: &[E],
     ) -> Vec<Vec<E>> {
         let v_table_col_builder = ChipletsVTableColBuilder::default();
diff --git a/processor/src/chiplets/mod.rs b/processor/src/chiplets/mod.rs
index 310d862728..c353d29d7d 100644
--- a/processor/src/chiplets/mod.rs
+++ b/processor/src/chiplets/mod.rs
@@ -1,9 +1,8 @@
 use crate::system::ContextId;
 
 use super::{
-    crypto::MerklePath, utils, BTreeMap, ChipletsTrace, ColMatrix, ExecutionError, Felt,
-    FieldElement, RangeChecker, StarkField, TraceFragment, Vec, Word, CHIPLETS_WIDTH, EMPTY_WORD,
-    ONE, ZERO,
+    crypto::MerklePath, utils, BTreeMap, ChipletsTrace, ExecutionError, Felt, FieldElement,
+    RangeChecker, StarkField, TraceFragment, Vec, Word, CHIPLETS_WIDTH, EMPTY_WORD, ONE, ZERO,
 };
 use miden_air::trace::chiplets::hasher::{Digest, HasherState};
 use vm_core::{code_blocks::OpBatch, Kernel};
diff --git a/processor/src/chiplets/tests.rs b/processor/src/chiplets/tests.rs
index aaa16d09b4..37fda7e045 100644
--- a/processor/src/chiplets/tests.rs
+++ b/processor/src/chiplets/tests.rs
@@ -1,6 +1,6 @@
 use crate::{
-    utils::get_trace_len, CodeBlock, DefaultHost, ExecutionOptions, ExecutionTrace, Kernel,
-    Operation, Process, StackInputs, Vec,
+    CodeBlock, DefaultHost, ExecutionOptions, ExecutionTrace, Kernel, Operation, Process,
+    StackInputs, Vec,
 };
 use miden_air::trace::{
     chiplets::{
@@ -117,11 +117,11 @@ fn build_trace(
     process.execute_code_block(&program, &CodeBlockTable::default()).unwrap();
 
     let (trace, _, _) = ExecutionTrace::test_finalize_trace(process);
-    let trace_len = get_trace_len(&trace) - ExecutionTrace::NUM_RAND_ROWS;
+    let trace_len = trace.num_rows() - ExecutionTrace::NUM_RAND_ROWS;
 
     (
-        trace[CHIPLETS_RANGE]
-            .to_vec()
+        trace
+            .get_column_range(CHIPLETS_RANGE)
             .try_into()
             .expect("failed to convert vector to array"),
         trace_len,
diff --git a/processor/src/decoder/auxiliary.rs b/processor/src/decoder/auxiliary.rs
index d42509f3c7..5e725a2de4 100644
--- a/processor/src/decoder/auxiliary.rs
+++ b/processor/src/decoder/auxiliary.rs
@@ -8,7 +8,6 @@ use miden_air::trace::{
 };
 
 use vm_core::{crypto::hash::RpoDigest, FieldElement, Operation};
-use winter_prover::matrix::ColMatrix;
 
 // CONSTANTS
 // ================================================================================================
@@ -39,7 +38,7 @@ impl AuxTraceBuilder {
     /// stack, block hash, and op group tables respectively.
     pub fn build_aux_columns<E: FieldElement<BaseField = Felt>>(
         &self,
-        main_trace: &ColMatrix<Felt>,
+        main_trace: &MainTrace,
         rand_elements: &[E],
     ) -> Vec<Vec<E>> {
         let block_stack_column_builder = BlockStackColumnBuilder::default();
diff --git a/processor/src/decoder/tests.rs b/processor/src/decoder/tests.rs
index 8586e44ea8..4772bfde2e 100644
--- a/processor/src/decoder/tests.rs
+++ b/processor/src/decoder/tests.rs
@@ -1,7 +1,6 @@
 use super::{
     super::{
-        utils::get_trace_len, ExecutionOptions, ExecutionTrace, Felt, Kernel, Operation, Process,
-        StackInputs, Word,
+        ExecutionOptions, ExecutionTrace, Felt, Kernel, Operation, Process, StackInputs, Word,
     },
     build_op_group,
 };
@@ -1202,11 +1201,11 @@ fn build_trace(stack_inputs: &[u64], program: &CodeBlock) -> (DecoderTrace, usiz
     process.execute_code_block(program, &CodeBlockTable::default()).unwrap();
 
     let (trace, _, _) = ExecutionTrace::test_finalize_trace(process);
-    let trace_len = get_trace_len(&trace) - ExecutionTrace::NUM_RAND_ROWS;
+    let trace_len = trace.num_rows() - ExecutionTrace::NUM_RAND_ROWS;
 
     (
-        trace[DECODER_TRACE_RANGE]
-            .to_vec()
+        trace
+            .get_column_range(DECODER_TRACE_RANGE)
             .try_into()
             .expect("failed to convert vector to array"),
         trace_len,
@@ -1230,11 +1229,11 @@ fn build_dyn_trace(
     process.execute_code_block(program, &cb_table).unwrap();
 
     let (trace, _, _) = ExecutionTrace::test_finalize_trace(process);
-    let trace_len = get_trace_len(&trace) - ExecutionTrace::NUM_RAND_ROWS;
+    let trace_len = trace.num_rows() - ExecutionTrace::NUM_RAND_ROWS;
 
     (
-        trace[DECODER_TRACE_RANGE]
-            .to_vec()
+        trace
+            .get_column_range(DECODER_TRACE_RANGE)
             .try_into()
             .expect("failed to convert vector to array"),
         trace_len,
@@ -1264,15 +1263,15 @@ fn build_call_trace(
     process.execute_code_block(program, &cb_table).unwrap();
 
     let (trace, _, _) = ExecutionTrace::test_finalize_trace(process);
-    let trace_len = get_trace_len(&trace) - ExecutionTrace::NUM_RAND_ROWS;
+    let trace_len = trace.num_rows() - ExecutionTrace::NUM_RAND_ROWS;
 
-    let sys_trace = trace[SYS_TRACE_RANGE]
-        .to_vec()
+    let sys_trace = trace
+        .get_column_range(SYS_TRACE_RANGE)
         .try_into()
         .expect("failed to convert vector to array");
 
-    let decoder_trace = trace[DECODER_TRACE_RANGE]
-        .to_vec()
+    let decoder_trace = trace
+        .get_column_range(DECODER_TRACE_RANGE)
         .try_into()
         .expect("failed to convert vector to array");
 
diff --git a/processor/src/range/aux_trace.rs b/processor/src/range/aux_trace.rs
index 531a6dba55..c2e634add3 100644
--- a/processor/src/range/aux_trace.rs
+++ b/processor/src/range/aux_trace.rs
@@ -1,4 +1,5 @@
-use super::{uninit_vector, BTreeMap, ColMatrix, Felt, FieldElement, Vec, NUM_RAND_ROWS};
+use super::{uninit_vector, BTreeMap, Felt, FieldElement, Vec, NUM_RAND_ROWS};
+use miden_air::trace::main_trace::MainTrace;
 use miden_air::trace::range::{M_COL_IDX, V_COL_IDX};
 use vm_core::StarkField;
 
@@ -42,7 +43,7 @@ impl AuxTraceBuilder {
     ///   requested by the Stack and Memory processors.
     pub fn build_aux_columns<E: FieldElement<BaseField = Felt>>(
         &self,
-        main_trace: &ColMatrix<Felt>,
+        main_trace: &MainTrace,
         rand_elements: &[E],
     ) -> Vec<Vec<E>> {
         let b_range = self.build_aux_col_b_range(main_trace, rand_elements);
@@ -53,7 +54,7 @@ impl AuxTraceBuilder {
     /// check lookups performed by user operations match those executed by the Range Checker.
     fn build_aux_col_b_range<E: FieldElement<BaseField = Felt>>(
         &self,
-        main_trace: &ColMatrix<Felt>,
+        main_trace: &MainTrace,
         rand_elements: &[E],
     ) -> Vec<E> {
         // run batch inversion on the lookup values
diff --git a/processor/src/range/mod.rs b/processor/src/range/mod.rs
index e1d12fe436..10f6e2c12c 100644
--- a/processor/src/range/mod.rs
+++ b/processor/src/range/mod.rs
@@ -1,6 +1,6 @@
 use super::{
-    trace::NUM_RAND_ROWS, utils::uninit_vector, BTreeMap, ColMatrix, Felt, FieldElement,
-    RangeCheckTrace, Vec, ZERO,
+    trace::NUM_RAND_ROWS, utils::uninit_vector, BTreeMap, Felt, FieldElement, RangeCheckTrace, Vec,
+    ZERO,
 };
 
 mod aux_trace;
diff --git a/processor/src/range/request.rs b/processor/src/range/request.rs
index 28cf9c0704..c71256eb2e 100644
--- a/processor/src/range/request.rs
+++ b/processor/src/range/request.rs
@@ -54,7 +54,7 @@ impl CycleRangeChecks {
     /// element in the field specified by E.
     pub fn to_stack_value<E: FieldElement<BaseField = Felt>>(
         &self,
-        main_trace: &ColMatrix<Felt>,
+        main_trace: &MainTrace,
         alphas: &[E],
     ) -> E {
         let mut value = E::ONE;
@@ -70,7 +70,7 @@ impl CycleRangeChecks {
     /// element in the field specified by E.
     fn to_mem_value<E: FieldElement<BaseField = Felt>>(
         &self,
-        main_trace: &ColMatrix<Felt>,
+        main_trace: &MainTrace,
         alphas: &[E],
     ) -> E {
         let mut value = E::ONE;
@@ -88,7 +88,7 @@ impl LookupTableRow for CycleRangeChecks {
     /// at least 1 alpha value. Includes all values included at this cycle from all processors.
     fn to_value<E: FieldElement<BaseField = Felt>>(
         &self,
-        main_trace: &ColMatrix<Felt>,
+        main_trace: &MainTrace,
         alphas: &[E],
     ) -> E {
         let stack_value = self.to_stack_value(main_trace, alphas);
@@ -115,7 +115,7 @@ impl LookupTableRow for RangeCheckRequest {
     /// at least 1 alpha value.
     fn to_value<E: FieldElement<BaseField = Felt>>(
         &self,
-        _main_trace: &ColMatrix<Felt>,
+        _main_trace: &MainTrace,
         alphas: &[E],
     ) -> E {
         let alpha: E = alphas[0];
diff --git a/processor/src/stack/aux_trace.rs b/processor/src/stack/aux_trace.rs
index ffb28ffb79..f0e6f590af 100644
--- a/processor/src/stack/aux_trace.rs
+++ b/processor/src/stack/aux_trace.rs
@@ -1,6 +1,6 @@
 use crate::trace::AuxColumnBuilder;
 
-use super::{ColMatrix, Felt, FieldElement, OverflowTableRow, Vec};
+use super::{Felt, FieldElement, OverflowTableRow, Vec};
 use miden_air::trace::main_trace::MainTrace;
 
 // AUXILIARY TRACE BUILDER
@@ -20,7 +20,7 @@ impl AuxTraceBuilder {
     /// column p1 describing states of the stack overflow table.
     pub fn build_aux_columns<E: FieldElement<BaseField = Felt>>(
         &self,
-        main_trace: &ColMatrix<Felt>,
+        main_trace: &MainTrace,
         rand_elements: &[E],
     ) -> Vec<Vec<E>> {
         let p1 = self.build_aux_column(main_trace, rand_elements);
diff --git a/processor/src/stack/mod.rs b/processor/src/stack/mod.rs
index aa0dfe6f51..aa2f42774d 100644
--- a/processor/src/stack/mod.rs
+++ b/processor/src/stack/mod.rs
@@ -1,6 +1,5 @@
 use super::{
-    BTreeMap, ColMatrix, Felt, FieldElement, StackInputs, StackOutputs, Vec, ONE,
-    STACK_TRACE_WIDTH, ZERO,
+    BTreeMap, Felt, FieldElement, StackInputs, StackOutputs, Vec, ONE, STACK_TRACE_WIDTH, ZERO,
 };
 use core::cmp;
 use vm_core::{stack::STACK_TOP_SIZE, Word, WORD_SIZE};
diff --git a/processor/src/trace/mod.rs b/processor/src/trace/mod.rs
index 00f95c8f1d..7dc889f433 100644
--- a/processor/src/trace/mod.rs
+++ b/processor/src/trace/mod.rs
@@ -5,6 +5,7 @@ use super::{
     stack::AuxTraceBuilder as StackAuxTraceBuilder, ColMatrix, Digest, Felt, FieldElement, Host,
     Process, StackTopState, Vec,
 };
+use miden_air::trace::main_trace::MainTrace;
 use miden_air::trace::{
     decoder::{NUM_USER_OP_HELPERS, USER_OP_HELPERS_OFFSET},
     AUX_TRACE_RAND_ELEMENTS, AUX_TRACE_WIDTH, DECODER_TRACE_OFFSET, MIN_TRACE_LEN,
@@ -50,7 +51,7 @@ pub struct AuxTraceBuilders {
 pub struct ExecutionTrace {
     meta: Vec<u8>,
     layout: TraceLayout,
-    main_trace: ColMatrix<Felt>,
+    main_trace: MainTrace,
     aux_trace_builders: AuxTraceBuilders,
     program_info: ProgramInfo,
     stack_outputs: StackOutputs,
@@ -86,8 +87,8 @@ impl ExecutionTrace {
         Self {
             meta: Vec::new(),
             layout: TraceLayout::new(TRACE_WIDTH, [AUX_TRACE_WIDTH], [AUX_TRACE_RAND_ELEMENTS]),
-            main_trace: ColMatrix::new(main_trace),
             aux_trace_builders: aux_trace_hints,
+            main_trace,
             program_info,
             stack_outputs,
             trace_len_summary,
@@ -173,7 +174,7 @@ impl ExecutionTrace {
     #[cfg(test)]
     pub fn test_finalize_trace<H>(
         process: Process<H>,
-    ) -> (Vec<Vec<Felt>>, AuxTraceBuilders, TraceLenSummary)
+    ) -> (MainTrace, AuxTraceBuilders, TraceLenSummary)
     where
         H: Host,
     {
@@ -276,7 +277,7 @@ impl Trace for ExecutionTrace {
 fn finalize_trace<H>(
     process: Process<H>,
     mut rng: RpoRandomCoin,
-) -> (Vec<Vec<Felt>>, AuxTraceBuilders, TraceLenSummary)
+) -> (MainTrace, AuxTraceBuilders, TraceLenSummary)
 where
     H: Host,
 {
@@ -341,5 +342,7 @@ where
         chiplets: chiplets_trace.aux_builder,
     };
 
-    (trace, aux_trace_hints, trace_len_summary)
+    let main_trace = MainTrace::new(ColMatrix::new(trace));
+
+    (main_trace, aux_trace_hints, trace_len_summary)
 }
diff --git a/processor/src/trace/tests/hasher.rs b/processor/src/trace/tests/hasher.rs
index c9a4c1bb44..8753e67a50 100644
--- a/processor/src/trace/tests/hasher.rs
+++ b/processor/src/trace/tests/hasher.rs
@@ -4,7 +4,8 @@ use super::{
     ZERO,
 };
 
-use crate::{ColMatrix, StackInputs};
+use crate::StackInputs;
+use miden_air::trace::main_trace::MainTrace;
 use miden_air::trace::{chiplets::hasher::P1_COL_IDX, AUX_TRACE_RAND_ELEMENTS};
 use vm_core::{
     crypto::merkle::{MerkleStore, MerkleTree, NodeIndex},
@@ -188,7 +189,7 @@ impl SiblingTableRow {
     /// at least 6 alpha values.
     pub fn to_value<E: FieldElement<BaseField = Felt>>(
         &self,
-        _main_trace: &ColMatrix<Felt>,
+        _main_trace: &MainTrace,
         alphas: &[E],
     ) -> E {
         // when the least significant bit of the index is 0, the sibling will be in the 3rd word
diff --git a/processor/src/trace/utils.rs b/processor/src/trace/utils.rs
index 00de7d3ea8..258eaaced0 100644
--- a/processor/src/trace/utils.rs
+++ b/processor/src/trace/utils.rs
@@ -6,7 +6,6 @@ use vm_core::{utils::uninit_vector, FieldElement};
 
 #[cfg(test)]
 use vm_core::{utils::ToElements, Operation};
-use winter_prover::matrix::ColMatrix;
 
 // TRACE FRAGMENT
 // ================================================================================================
@@ -221,19 +220,18 @@ pub trait AuxColumnBuilder<E: FieldElement<BaseField = Felt>> {
     }
 
     /// Builds the chiplets bus auxiliary trace column.
-    fn build_aux_column(&self, main_trace: &ColMatrix<Felt>, alphas: &[E]) -> Vec<E> {
-        let main_trace = MainTrace::new(main_trace);
+    fn build_aux_column(&self, main_trace: &MainTrace, alphas: &[E]) -> Vec<E> {
         let mut responses_prod: Vec<E> = unsafe { uninit_vector(main_trace.num_rows()) };
         let mut requests: Vec<E> = unsafe { uninit_vector(main_trace.num_rows()) };
 
-        responses_prod[0] = self.init_responses(&main_trace, alphas);
-        requests[0] = self.init_requests(&main_trace, alphas);
+        responses_prod[0] = self.init_responses(main_trace, alphas);
+        requests[0] = self.init_requests(main_trace, alphas);
 
         let mut requests_running_prod = E::ONE;
         for row_idx in 0..main_trace.num_rows() - 1 {
             responses_prod[row_idx + 1] =
-                responses_prod[row_idx] * self.get_responses_at(&main_trace, alphas, row_idx);
-            requests[row_idx + 1] = self.get_requests_at(&main_trace, alphas, row_idx);
+                responses_prod[row_idx] * self.get_responses_at(main_trace, alphas, row_idx);
+            requests[row_idx + 1] = self.get_requests_at(main_trace, alphas, row_idx);
             requests_running_prod *= requests[row_idx + 1];
         }
 
diff --git a/prover/src/gpu.rs b/prover/src/gpu.rs
index af04c4cee3..51a0a16abb 100644
--- a/prover/src/gpu.rs
+++ b/prover/src/gpu.rs
@@ -8,6 +8,7 @@ use super::{
     ExecutionProver, ExecutionTrace, Felt, FieldElement, Level, ProcessorAir, PublicInputs,
     WinterProofOptions,
 };
+use air::trace::main_trace::MainTrace;
 use elsa::FrozenVec;
 use ministark_gpu::{
     plan::{gen_rpo_merkle_tree, GpuRpo256RowMajor},
@@ -62,7 +63,7 @@ where
     fn new_trace_lde<E: FieldElement<BaseField = Felt>>(
         &self,
         trace_info: &TraceInfo,
-        main_trace: &ColMatrix<Felt>,
+        main_trace: &MainTrace,
         domain: &StarkDomain<Felt>,
     ) -> (Self::TraceLde<E>, TracePolyTable<E>) {
         MetalRpoTraceLde::new(trace_info, main_trace, domain)
@@ -200,7 +201,7 @@ impl<E: FieldElement<BaseField = Felt>> MetalRpoTraceLde<E> {
     /// segment and the new [DefaultTraceLde].
     pub fn new(
         trace_info: &TraceInfo,
-        main_trace: &ColMatrix<Felt>,
+        main_trace: &MainTrace,
         domain: &StarkDomain<Felt>,
     ) -> (Self, TracePolyTable<E>) {
         // extend the main execution trace and build a Merkle tree from the extended trace