diff --git a/bin/README.md b/bin/README.md
index 759fae9a..3fb5b6e5 100644
--- a/bin/README.md
+++ b/bin/README.md
@@ -4,6 +4,8 @@
 * faster than browsers
 * early beta, if something breaks just restart it
 
+
+
 # [`copyparty-fuse.py`](copyparty-fuse.py)
 * mount a copyparty server as a local filesystem (read-only)
 * **supports Windows!** -- expect `194 MiB/s` sequential read
@@ -53,6 +55,7 @@ you could replace winfsp with [dokan](https://github.com/dokan-dev/dokany/releas
 * copyparty can Popen programs like these during file indexing to collect additional metadata
 
 
+
 # [`dbtool.py`](dbtool.py)
 upgrade utility which can show db info and help transfer data between databases, for example when a new version of copyparty is incompatible with the old DB and automatically rebuilds the DB from scratch, but you have some really expensive `-mtp` parsers and want to copy over the tags from the old db
 
@@ -69,6 +72,7 @@ cd /mnt/nas/music/.hist
 ```
 
 
+
 # [`prisonparty.sh`](prisonparty.sh)
 * run copyparty in a chroot, preventing any accidental file access
 * creates bindmounts for /bin, /lib, and so on, see `sysdirs=`
diff --git a/bin/mtag/README.md b/bin/mtag/README.md
index d4603b9c..2adda0ba 100644
--- a/bin/mtag/README.md
+++ b/bin/mtag/README.md
@@ -10,6 +10,7 @@ some of these rely on libraries which are not MIT-compatible
 
 these do not have any problematic dependencies:
 
+* [cksum.py](./cksum.py) computes various checksums
 * [exe.py](./exe.py) grabs metadata from .exe and .dll files (example for retrieving multiple tags with one parser)
 * [wget.py](./wget.py) lets you download files by POSTing URLs to copyparty
 
diff --git a/bin/mtag/cksum.py b/bin/mtag/cksum.py
new file mode 100755
index 00000000..c70a1b87
--- /dev/null
+++ b/bin/mtag/cksum.py
@@ -0,0 +1,89 @@
+#!/usr/bin/env python3
+
+import sys
+import json
+import zlib
+import struct
+import base64
+import hashlib
+
+try:
+    from copyparty.util import fsenc
+except:
+
+    def fsenc(p):
+        return p
+
+
+"""
+calculates various checksums for uploads,
+usage: -mtp crc32,md5,sha1,sha256b=bin/mtag/cksum.py
+"""
+
+
+def main():
+    config = "crc32 md5 md5b sha1 sha1b sha256 sha256b sha512/240 sha512b/240"
+    # b suffix = base64 encoded
+    # slash = truncate to n bits
+
+    known = {
+        "md5": hashlib.md5,
+        "sha1": hashlib.sha1,
+        "sha256": hashlib.sha256,
+        "sha512": hashlib.sha512,
+    }
+    config = config.split()
+    hashers = {
+        k: v()
+        for k, v in known.items()
+        if k in [x.split("/")[0].rstrip("b") for x in known]
+    }
+    crc32 = 0 if "crc32" in config else None
+
+    with open(fsenc(sys.argv[1]), "rb", 512 * 1024) as f:
+        while True:
+            buf = f.read(64 * 1024)
+            if not buf:
+                break
+
+            for x in hashers.values():
+                x.update(buf)
+
+            if crc32 is not None:
+                crc32 = zlib.crc32(buf, crc32)
+
+    ret = {}
+    for s in config:
+        alg = s.split("/")[0]
+        b64 = alg.endswith("b")
+        alg = alg.rstrip("b")
+        if alg in hashers:
+            v = hashers[alg].digest()
+        elif alg == "crc32":
+            v = crc32
+            if v < 0:
+                v &= 2 ** 32 - 1
+            v = struct.pack(">L", v)
+        else:
+            raise Exception("what is {}".format(s))
+
+        if "/" in s:
+            v = v[: int(int(s.split("/")[1]) / 8)]
+
+        if b64:
+            v = base64.b64encode(v).decode("ascii").rstrip("=")
+        else:
+            try:
+                v = v.hex()
+            except:
+                import binascii
+
+                v = binascii.hexlify(v)
+
+        ret[s] = v
+
+    print(json.dumps(ret, indent=4))
+
+
+if __name__ == "__main__":
+    main()