Can I decode Protobuf without the .proto file?

Yes. The wire format carries field numbers and wire types, so the structure can be fully recovered — just not the names or exact types. A varint could be an int, bool, enum or zigzag-encoded sint, and a length-delimited field could be a string, bytes, a nested message or a packed array. The tool shows the most likely reading and lists alternatives in grey. Paste the .proto for exact results.

How do I decode captured gRPC traffic?

Each gRPC message starts with a 5-byte frame header: a 1-byte compression flag and a 4-byte big-endian length. Strip those 5 bytes before pasting; if the flag is 1 the payload is also gzip-compressed and must be decompressed first. gRPC-Web text is the same after Base64 decoding.

Why can the same bytes be detected as different formats?

None of these formats has a magic number, and short inputs are often valid Protobuf, MessagePack and CBOR at the same time. Auto-detection prefers the one that decodes completely into the most sensible structure (BSON is easy thanks to its length header). If it guesses wrong, choose the format manually.

How are big integers and binary data shown in JSON?

Integers beyond JavaScript’s safe range (2^53) are output as strings so no precision is lost, and bytes are output as Base64 strings, matching Protobuf’s JSON mapping. BSON uses MongoDB Extended JSON — {"$oid": "…"}, {"$date": …}, {"$numberLong": "…"} — so types survive encoding back to BSON.

💾

💾Protobuf Decoder Online - MessagePack, CBOR & BSON Viewer

Captured some gRPC / Protobuf bytes, found MessagePack in Redis, got CBOR from a device or a BSON dump from MongoDB? Paste it as hex (spaces, 0x and \x styles all work) or Base64, or open the file, and the format is detected for you. Without a schema, Protobuf is decoded straight from the wire format: field numbers and wire types, varints read as unsigned, signed int64 and zigzag (sint), fixed32/64 as integers and floats, and length-delimited fields tried as printable string, nested message and packed repeated values before falling back to bytes — with the other possible readings listed alongside, plus protoc --decode_raw-style text output. Paste a .proto file to decode with real field and enum names, including nested types, repeated, packed, map and oneof fields. MessagePack covers every type plus the timestamp extension; CBOR follows RFC 8949 with tags (dates, bignums, UUIDs, embedded CBOR), half-precision floats and indefinite-length strings, arrays and maps, and produces diagnostic notation; BSON handles ObjectId, dates, Decimal128 and binary and outputs Extended JSON. Results appear as a tree and as JSON: hover a node to highlight its bytes in the hex view, or click a byte to jump to its node; decoding errors point at the exact offset. It also works in reverse, encoding JSON to MessagePack, CBOR or BSON. Everything runs locally in your browser.

Free · No sign-up · Runs in your browserLast updated: 2026-10-26

Loading tool...

How to use Protobuf, MessagePack, CBOR & BSON Decoder

1Paste hex or Base64, or open a binary file
2Pick the format (auto-detected by default); for Protobuf optionally paste a .proto and choose the root message
3Explore the tree alongside the hex view and copy the JSON

FAQ

Can I decode Protobuf without the .proto file?: Yes. The wire format carries field numbers and wire types, so the structure can be fully recovered — just not the names or exact types. A varint could be an int, bool, enum or zigzag-encoded sint, and a length-delimited field could be a string, bytes, a nested message or a packed array. The tool shows the most likely reading and lists alternatives in grey. Paste the .proto for exact results.
How do I decode captured gRPC traffic?: Each gRPC message starts with a 5-byte frame header: a 1-byte compression flag and a 4-byte big-endian length. Strip those 5 bytes before pasting; if the flag is 1 the payload is also gzip-compressed and must be decompressed first. gRPC-Web text is the same after Base64 decoding.
Why can the same bytes be detected as different formats?: None of these formats has a magic number, and short inputs are often valid Protobuf, MessagePack and CBOR at the same time. Auto-detection prefers the one that decodes completely into the most sensible structure (BSON is easy thanks to its length header). If it guesses wrong, choose the format manually.
How are big integers and binary data shown in JSON?: Integers beyond JavaScript’s safe range (2^53) are output as strings so no precision is lost, and bytes are output as Base64 strings, matching Protobuf’s JSON mapping. BSON uses MongoDB Extended JSON — {"$oid": "…"}, {"$date": …}, {"$numberLong": "…"} — so types survive encoding back to BSON.