diff --git a/spec_test.go b/spec_test.go
new file mode 100644
index 0000000..6e56898
--- /dev/null
+++ b/spec_test.go
@@ -0,0 +1,147 @@
+package multibase
+
+import (
+	"encoding/csv"
+	"os"
+	"path/filepath"
+	"strconv"
+	"strings"
+	"testing"
+	"unicode/utf8"
+)
+
+func TestSpec(t *testing.T) {
+	file, err := os.Open("spec/multibase.csv")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer file.Close()
+
+	reader := csv.NewReader(file)
+	reader.LazyQuotes = false
+	reader.FieldsPerRecord = 3
+	reader.TrimLeadingSpace = true
+
+	values, err := reader.ReadAll()
+	if err != nil {
+		t.Error(err)
+	}
+	expectedEncodings := make(map[Encoding]string, len(values)-1)
+	for _, v := range values[1:] {
+		encoding := v[0]
+		codeStr := v[1]
+
+		var code Encoding
+		if strings.HasPrefix(codeStr, "0x") {
+			i, err := strconv.ParseUint(codeStr[2:], 16, 64)
+			if err != nil {
+				t.Errorf("invalid multibase byte %q", codeStr)
+				continue
+			}
+			code = Encoding(i)
+		} else {
+			codeRune, length := utf8.DecodeRuneInString(codeStr)
+			if code == utf8.RuneError {
+				t.Errorf("multibase %q wasn't valid utf8", codeStr)
+				continue
+			}
+			if length != len(codeStr) {
+				t.Errorf("multibase %q wasn't a single character", codeStr)
+				continue
+			}
+			code = Encoding(codeRune)
+		}
+		expectedEncodings[code] = encoding
+	}
+
+	for name, enc := range Encodings {
+		expectedName, ok := expectedEncodings[enc]
+		if !ok {
+			t.Errorf("encoding %q (%c) not defined in the spec", name, enc)
+			continue
+		}
+		if expectedName != name {
+			t.Errorf("encoding %q (%c) has unexpected name %q", expectedName, enc, name)
+		}
+	}
+}
+func TestSpecVectors(t *testing.T) {
+	files, err := filepath.Glob("spec/tests/test[0-9]*.csv")
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, fname := range files {
+		t.Run(fname, func(t *testing.T) {
+			file, err := os.Open(fname)
+			if err != nil {
+				t.Error(err)
+				return
+			}
+			defer file.Close()
+			reader := csv.NewReader(file)
+			reader.LazyQuotes = false
+			reader.FieldsPerRecord = 2
+			reader.TrimLeadingSpace = true
+
+			values, err := reader.ReadAll()
+			if err != nil {
+				t.Error(err)
+			}
+			if len(values) == 0 {
+				t.Error("no test values")
+				return
+			}
+			header := values[0]
+
+			var decodeOnly bool
+			switch header[0] {
+			case "encoding":
+			case "non-canonical encoding":
+				decodeOnly = true
+			default:
+				t.Errorf("invalid test spec %q", fname)
+				return
+			}
+
+			testValue, err := strconv.Unquote("\"" + header[1] + "\"")
+			if err != nil {
+				t.Error("failed to unquote testcase:", err)
+				return
+			}
+
+			for _, testCase := range values[1:] {
+				encodingName := testCase[0]
+				expected := testCase[1]
+
+				t.Run(encodingName, func(t *testing.T) {
+					encoder, err := EncoderByName(encodingName)
+					if err != nil {
+						t.Skipf("skipping %s: not supported", encodingName)
+						return
+					}
+					if !decodeOnly {
+						t.Logf("encoding %q with %s", testValue, encodingName)
+						actual := encoder.Encode([]byte(testValue))
+						if expected != actual {
+							t.Errorf("expected %q, got %q", expected, actual)
+						}
+					}
+					t.Logf("decoding %q", expected)
+					encoding, decoded, err := Decode(expected)
+					if err != nil {
+						t.Error("failed to decode:", err)
+						return
+					}
+					expectedEncoding := Encodings[encodingName]
+					if encoding != expectedEncoding {
+						t.Errorf("expected encoding to be %c, got %c", expectedEncoding, encoding)
+					}
+					if string(decoded) != testValue {
+						t.Errorf("failed to decode %q to %q, got %q", expected, testValue, string(decoded))
+					}
+				})
+
+			}
+		})
+	}
+}