diff --git a/node/core/batch.go b/node/core/batch.go
index 987c4bf0..9c851956 100644
--- a/node/core/batch.go
+++ b/node/core/batch.go
@@ -178,6 +178,17 @@ func (e *Executor) CalculateCapWithProposalBlock(currentBlockBytes []byte, curre
 		return false, err
 	}
 
+	// MPT fork: force batch points on the 1st and 2nd post-fork blocks, so the 1st post-fork block
+	// becomes a single-block batch: [H1, H2).
+	force, err := e.forceBatchPointForMPTFork(height, block.Timestamp, block.StateRoot, block.Hash)
+	if err != nil {
+		return false, err
+	}
+	if force {
+		e.logger.Info("MPT fork: force batch point", "height", height, "timestamp", block.Timestamp)
+		return true, nil
+	}
+
 	var exceeded bool
 	if e.isBatchUpgraded(block.Timestamp) {
 		exceeded, err = e.batchingCache.batchData.WillExceedCompressedSizeLimit(e.batchingCache.currentBlockContext, e.batchingCache.currentTxsPayload)
@@ -187,6 +198,79 @@ func (e *Executor) CalculateCapWithProposalBlock(currentBlockBytes []byte, curre
 	return exceeded, err
 }
 
+// forceBatchPointForMPTFork forces batch points at the 1st and 2nd block after the MPT fork time.
+//
+// Design goals:
+// - Minimal change: only affects batch-point decision logic.
+// - Stability: CalculateCapWithProposalBlock can be called multiple times at the same height; return must be consistent.
+// - Performance: after handling (or skipping beyond) the fork boundary, no more HeaderByNumber calls are made.
+func (e *Executor) forceBatchPointForMPTFork(height uint64, blockTime uint64, stateRoot common.Hash, blockHash common.Hash) (bool, error) {
+	// If we already decided to force at this height, keep returning true without extra RPCs.
+	if e.mptForkForceHeight == height && height != 0 {
+		return true, nil
+	}
+	// If fork boundary is already handled and this isn't a forced height, fast exit.
+	if e.mptForkStage >= 2 {
+		return false, nil
+	}
+
+	// Ensure we have fork time cached (0 means disabled).
+	if e.mptForkTime == 0 {
+		e.mptForkTime = e.l2Client.MPTForkTime()
+	}
+	forkTime := e.mptForkTime
+	if forkTime == 0 || blockTime < forkTime {
+		return false, nil
+	}
+	if height == 0 {
+		return false, nil
+	}
+
+	// Check parent block time to detect the 1st post-fork block (H1).
+	parent, err := e.l2Client.HeaderByNumber(context.Background(), big.NewInt(int64(height-1)))
+	if err != nil {
+		return false, err
+	}
+	if parent.Time < forkTime {
+		// Log H1 (the 1st post-fork block) state root
+		// This stateRoot is intended to be used as the Rollup contract "genesis state root"
+		// when we reset/re-initialize the genesis state root during the MPT upgrade.
+		e.logger.Info(
+			"MPT_FORK_H1_GENESIS_STATE_ROOT",
+			"height", height,
+			"timestamp", blockTime,
+			"forkTime", forkTime,
+			"stateRoot", stateRoot.Hex(),
+			"blockHash", blockHash.Hex(),
+		)
+		e.mptForkStage = 1
+		e.mptForkForceHeight = height
+		return true, nil
+	}
+
+	// If parent is already post-fork, we may be at the 2nd post-fork block (H2) or later.
+	if height < 2 {
+		// We cannot be H2; mark done to avoid future calls.
+		e.mptForkStage = 2
+		return false, nil
+	}
+
+	grandParent, err := e.l2Client.HeaderByNumber(context.Background(), big.NewInt(int64(height-2)))
+	if err != nil {
+		return false, err
+	}
+	if grandParent.Time < forkTime {
+		// This is H2 (2nd post-fork block).
+		e.mptForkStage = 2
+		e.mptForkForceHeight = height
+		return true, nil
+	}
+
+	// Beyond H2: nothing to do (can't retroactively fix). Mark done for performance.
+	e.mptForkStage = 2
+	return false, nil
+}
+
 func (e *Executor) AppendBlsData(height int64, batchHash []byte, data l2node.BlsData) error {
 	if len(batchHash) != 32 {
 		return fmt.Errorf("wrong batchHash length. expected: 32, actual: %d", len(batchHash))
diff --git a/node/core/executor.go b/node/core/executor.go
index 2757600f..0f16ca00 100644
--- a/node/core/executor.go
+++ b/node/core/executor.go
@@ -56,6 +56,13 @@ type Executor struct {
 	rollupABI             *abi.ABI
 	batchingCache         *BatchingCache
 
+	// MPT fork handling: force batch points at the 1st and 2nd block after fork.
+	// This state machine exists to avoid repeated HeaderByNumber calls after the fork is handled,
+	// while keeping results stable if CalculateCapWithProposalBlock is called multiple times at the same height.
+	mptForkTime        uint64 // cached from geth eth_config.morph.mptForkTime (0 means disabled/unknown)
+	mptForkStage       uint8  // 0: not handled, 1: forced H1, 2: done (forced H2 or skipped beyond H2)
+	mptForkForceHeight uint64 // if equals current height, must return true (stability across multiple calls)
+
 	logger  tmlog.Logger
 	metrics *Metrics
 }
@@ -148,6 +155,7 @@ func NewExecutor(newSyncFunc NewSyncerFunc, config *Config, tmPubKey crypto.PubK
 		batchingCache:         NewBatchingCache(),
 		UpgradeBatchTime:      config.UpgradeBatchTime,
 		blsKeyCheckForkHeight: config.BlsKeyCheckForkHeight,
+		mptForkTime:           l2Client.MPTForkTime(),
 		logger:                logger,
 		metrics:               PrometheusMetrics("morphnode"),
 	}
diff --git a/node/types/retryable_client.go b/node/types/retryable_client.go
index 53d94f35..6f954156 100644
--- a/node/types/retryable_client.go
+++ b/node/types/retryable_client.go
@@ -122,6 +122,13 @@ type RetryableClient struct {
 	logger         tmlog.Logger
 }
 
+// MPTForkTime returns the configured MPT fork/switch timestamp fetched from geth (eth_config).
+// Note: this is a local value stored in the client; it does not perform any RPC.
+func (rc *RetryableClient) MPTForkTime() uint64 {
+	return rc.switchTime
+}
+
+
 // NewRetryableClient creates a new retryable client with the given switch time.
 // Will retry calling the api, if the connection is refused.
 //