diff --git a/README.md b/README.md
new file mode 100644
index 0000000..55cebcb
--- /dev/null
+++ b/README.md
@@ -0,0 +1,120 @@
+6502 inflate routine
+====================
+
+[DEFLATE](http://en.wikipedia.org/wiki/DEFLATE) is a popular compression format,
+used in ZIP, gzip, PNG and many other formats.
+
+In 2001 I wrote a DEFLATE decompression routine (called "inflate")
+in the [6502](http://en.wikipedia.org/wiki/6502) assembly language.
+In 2007 I optimized it so it is about 30% shorter and 10% faster than before.
+
+Compilation
+-----------
+
+The routine uses three memory areas:
+
+* `inflate` - code and initialized data (509 bytes)
+* `inflate_data` - uninitialized data (764 bytes)
+* `inflate_zp` - variables on zero page
+
+You must select these locations at compile time, for example:
+
+    xasm -d inflate="$b700" -d inflate_data="$b900" -d inflate_zp="$f0" inflate.asx
+
+Source code uses [xasm](https://github.com/pfusik/xasm) syntax.
+This cross-assembler includes many original syntax extensions.
+
+Usage
+-----
+
+The `inflate` routine assumes that the compressed and uncompressed data fit
+in the memory. Before calling `inflate` set the locations of compressed
+and uncompressed data in zero-page variables:
+
+    mwa #compressedData inflate_zp
+    mwa #uncompressedData inflate_zp+2
+    jsr inflate
+
+As the compressed data is read sequentially and only once, it is possible
+to overlap the compressed and uncompressed data. That is, the data being
+uncompressed can be stored in place of some compressed data which has been
+already read.
+
+It is also possible to get the compressed data from any forward-only stream.
+In this case, modify the `getBit` routine to look like this:
+
+    getBit
+        lsr getBit_buffer
+        bne getBit_return
+        pha
+        stx getBit_buffer
+        jsr readByte
+        ldx getBit_buffer
+        ldy #0
+        sec
+        ror @
+        sta getBit_buffer
+        pla
+    getBit_return
+        rts
+
+Compression
+-----------
+
+There are several ways to get DEFLATE compressed data.
+Originally I wrote a command-line utility called `deflater`
+which used the standard [zlib library](http://www.zlib.net/).
+However, better compression can be obtained with [7-Zip](http://7-zip.org/).
+It supports the gzip format which is a thin layer on top of DEFLATE.
+My program `gzip2deflate` extracts the DEFLATE stream from a gzip file.
+It reads gzip on standard input and writer DEFLATE to standard output.
+I recommend using it like this:
+
+    7z a -tgzip -mx=9 -so dummy INPUT_FILE | gzip2deflate >OUTPUT_FILE.dfl
+
+If you don't have 7-Zip, use:
+
+    gzip -c -9 INPUT_FILE | gzip2deflate >OUTPUT_FILE.dfl
+
+If you are looking for maximum compression, [KZIP](http://advsys.net/ken/utils.htm)
+is also worth a try. It creates ZIP files, so I wrote `zip2deflate` to extract
+DEFLATE data from ZIP.
+
+Windows binaries of `gzip2deflate` and `zip2deflate` are
+[available for download](http://pfusik.github.io/zlib6502/2deflate.zip).
+For other platforms you will need to compile these programs yourself.
+
+zlib?
+-----
+
+This project is named zlib6502, but only supports DEFLATE decompression.
+Someday I'm going to include more functions, including compression.
+
+Meanwhile, you may look at [cc65](https://github.com/oliverschmidt/cc65) `zlib.h`.
+This is my old code, which includes an old version of `inflate`
+plus zlib-compatible `uncompress`, `adler32` and `crc32`.
+
+License
+-------
+
+This code is licensed under the standard zlib license.
+
+Copyright (C) 2001-2013 Piotr Fusik
+
+This software is provided 'as-is', without any express or implied
+warranty.  In no event will the authors be held liable for any damages
+arising from the use of this software.
+
+Permission is granted to anyone to use this software for any purpose,
+including commercial applications, and to alter it and redistribute it
+freely, subject to the following restrictions:
+
+1. The origin of this software must not be misrepresented; you must not
+   claim that you wrote the original software. If you use this software
+   in a product, an acknowledgment in the product documentation would be
+   appreciated but is not required.
+
+2. Altered source versions must be plainly marked as such, and must not be
+   misrepresented as being the original software.
+
+3. This notice may not be removed or altered from any source distribution.