diff --git a/codec/dagcbor/marshal.go b/codec/dagcbor/marshal.go
index ee8a8b38..6798d050 100644
--- a/codec/dagcbor/marshal.go
+++ b/codec/dagcbor/marshal.go
@@ -245,3 +245,125 @@ func marshalMap(n datamodel.Node, tk *tok.Token, sink shared.TokenSink, options
 	_, err := sink.Step(tk)
 	return err
 }
+
+// EncodedLength will calculate the length in bytes that the encoded form of the
+// provided Node will occupy.
+//
+// Note that this function requires a full walk of the Node's graph, which may
+// not necessarily be a trivial cost and will incur some allocations. Using this
+// method to calculate buffers to pre-allocate may not result in performance
+// gains, but rather incur an overall cost. Use with care.
+func EncodedLength(n datamodel.Node) (int64, error) {
+	switch n.Kind() {
+	case datamodel.Kind_Invalid:
+		return 0, fmt.Errorf("cannot traverse a node that is absent")
+	case datamodel.Kind_Null:
+		return 1, nil // 0xf6
+	case datamodel.Kind_Map:
+		length := uintLength(uint64(n.Length())) // length prefixed major 5
+		for itr := n.MapIterator(); !itr.Done(); {
+			k, v, err := itr.Next()
+			if err != nil {
+				return 0, err
+			}
+			keyLength, err := EncodedLength(k)
+			if err != nil {
+				return 0, err
+			}
+			length += keyLength
+			valueLength, err := EncodedLength(v)
+			if err != nil {
+				return 0, err
+			}
+			length += valueLength
+		}
+		return length, nil
+	case datamodel.Kind_List:
+		nl := n.Length()
+		length := uintLength(uint64(nl)) // length prefixed major 4
+		for i := int64(0); i < nl; i++ {
+			v, err := n.LookupByIndex(i)
+			if err != nil {
+				return 0, err
+			}
+			innerLength, err := EncodedLength(v)
+			if err != nil {
+				return 0, err
+			}
+			length += innerLength
+		}
+		return length, nil
+	case datamodel.Kind_Bool:
+		return 1, nil // 0xf4 or 0xf5
+	case datamodel.Kind_Int:
+		v, err := n.AsInt()
+		if err != nil {
+			return 0, err
+		}
+		if v < 0 {
+			v = -v - 1 // negint is stored as one less than actual
+		}
+		return uintLength(uint64(v)), nil // major 0 or 1, as small as possible
+	case datamodel.Kind_Float:
+		return 9, nil // always major 7 and 64-bit float
+	case datamodel.Kind_String:
+		v, err := n.AsString()
+		if err != nil {
+			return 0, err
+		}
+
+		return uintLength(uint64(len(v))) + int64(len(v)), nil // length prefixed major 3
+	case datamodel.Kind_Bytes:
+		v, err := n.AsBytes()
+		if err != nil {
+			return 0, err
+		}
+		return uintLength(uint64(len(v))) + int64(len(v)), nil // length prefixed major 2
+	case datamodel.Kind_Link:
+		v, err := n.AsLink()
+		if err != nil {
+			return 0, err
+		}
+		switch lnk := v.(type) {
+		case cidlink.Link:
+			length := int64(2)                    // tag,42: 0xd82a
+			bl := int64(len(lnk.Bytes())) + 1     // additional 0x00 in front of the CID bytes
+			length += uintLength(uint64(bl)) + bl // length prefixed major 2
+			return length, err
+		default:
+			return 0, fmt.Errorf("schemafree link emission only supported by this codec for CID type links")
+		}
+	default:
+		panic("unreachable")
+	}
+}
+
+// Calculate how many bytes an integer, and therefore also the leading bytes of
+// a length-prefixed token. CBOR will pack it up into the smallest possible
+// uint representation, even merging it with the major if it's <=23.
+
+type boundaryLength struct {
+	upperBound uint64
+	length     int64
+}
+
+var lengthBoundaries = []boundaryLength{
+	{24, 1},         // packed major|minor
+	{256, 2},        // major, 8-bit length
+	{65536, 3},      // major, 16-bit length
+	{4294967296, 5}, // major, 32-bit length
+	{0, 9},          // major, 64-bit length
+}
+
+func uintLength(ii uint64) int64 {
+	for _, lb := range lengthBoundaries {
+		if ii < lb.upperBound {
+			return lb.length
+		}
+	}
+	// maximum number of bytes to pack this int
+	// if this int is used as a length prefix for a map, list, string or bytes
+	// then we likely have a very bad Node that shouldn't be encoded, but the
+	// encoder may raise problems with that if the memory allocator doesn't first.
+	return lengthBoundaries[len(lengthBoundaries)-1].length
+}
diff --git a/codec/dagcbor/marshal_test.go b/codec/dagcbor/marshal_test.go
new file mode 100644
index 00000000..50eacf6c
--- /dev/null
+++ b/codec/dagcbor/marshal_test.go
@@ -0,0 +1,70 @@
+package dagcbor
+
+import (
+	"bytes"
+	"math/rand"
+	"testing"
+	"time"
+
+	qt "github.com/frankban/quicktest"
+	"github.com/ipld/go-ipld-prime/datamodel"
+	basicnode "github.com/ipld/go-ipld-prime/node/basic"
+	"github.com/ipld/go-ipld-prime/testutil/garbage"
+)
+
+func calculateActualLength(t *testing.T, n datamodel.Node) int64 {
+	var buf bytes.Buffer
+	err := Encode(n, &buf)
+	qt.Assert(t, err, qt.IsNil)
+	return int64(buf.Len())
+}
+
+func verifyEstimatedSize(t *testing.T, n datamodel.Node) {
+	estimatedLength, err := EncodedLength(n)
+	qt.Assert(t, err, qt.IsNil)
+	actualLength := calculateActualLength(t, n)
+	qt.Assert(t, estimatedLength, qt.Equals, actualLength)
+}
+
+func TestEncodedLength(t *testing.T) {
+	t.Run("int boundaries", func(t *testing.T) {
+		for ii := 0; ii < 4; ii++ {
+			verifyEstimatedSize(t, basicnode.NewInt(int64(lengthBoundaries[ii].upperBound)))
+			verifyEstimatedSize(t, basicnode.NewInt(int64(lengthBoundaries[ii].upperBound)-1))
+			verifyEstimatedSize(t, basicnode.NewInt(int64(lengthBoundaries[ii].upperBound)+1))
+			verifyEstimatedSize(t, basicnode.NewInt(-1*int64(lengthBoundaries[ii].upperBound)))
+			verifyEstimatedSize(t, basicnode.NewInt(-1*int64(lengthBoundaries[ii].upperBound)-1))
+			verifyEstimatedSize(t, basicnode.NewInt(-1*int64(lengthBoundaries[ii].upperBound)+1))
+		}
+	})
+
+	t.Run("small garbage", func(t *testing.T) {
+		seed := time.Now().Unix()
+		t.Logf("randomness seed: %v\n", seed)
+		rnd := rand.New(rand.NewSource(seed))
+		for i := 0; i < 1000; i++ {
+			gbg := garbage.Generate(rnd, garbage.TargetBlockSize(1<<6))
+			verifyEstimatedSize(t, gbg)
+		}
+	})
+
+	t.Run("medium garbage", func(t *testing.T) {
+		seed := time.Now().Unix()
+		t.Logf("randomness seed: %v\n", seed)
+		rnd := rand.New(rand.NewSource(seed))
+		for i := 0; i < 100; i++ {
+			gbg := garbage.Generate(rnd, garbage.TargetBlockSize(1<<16))
+			verifyEstimatedSize(t, gbg)
+		}
+	})
+
+	t.Run("large garbage", func(t *testing.T) {
+		seed := time.Now().Unix()
+		t.Logf("randomness seed: %v\n", seed)
+		rnd := rand.New(rand.NewSource(seed))
+		for i := 0; i < 10; i++ {
+			gbg := garbage.Generate(rnd, garbage.TargetBlockSize(1<<20))
+			verifyEstimatedSize(t, gbg)
+		}
+	})
+}
diff --git a/codec/dagcbor/roundtrip_test.go b/codec/dagcbor/roundtrip_test.go
index fef0f5a6..babce5e5 100644
--- a/codec/dagcbor/roundtrip_test.go
+++ b/codec/dagcbor/roundtrip_test.go
@@ -56,6 +56,11 @@ func TestRoundtrip(t *testing.T) {
 		qt.Assert(t, err, qt.IsNil)
 		qt.Check(t, buf.String(), qt.Equals, serial)
 	})
+	t.Run("length", func(t *testing.T) {
+		length, err := EncodedLength(n)
+		qt.Assert(t, err, qt.IsNil)
+		qt.Check(t, length, qt.Equals, int64(len(serial)))
+	})
 	t.Run("decoding", func(t *testing.T) {
 		buf := strings.NewReader(serial)
 		nb := basicnode.Prototype.Map.NewBuilder()