The Pappy (**P**roxy **A**ttack **P**roxy **P**rox**Y**) Proxy is an intercepting proxy for performing web application security testing. Its features are often similar, or straight up rippoffs from [Burp Suite](https://portswigger.net/burp/). However, Burp Suite is neither open source nor a command line tool, thus making a proxy like Pappy inevitable. The project is still in its early stages, so there are bugs and only the bare minimum features, but it can already do some cool stuff.
**I am taking any and all feature requests.** If you've used Burp and had any inconvenience with it, tell me about it and I'll do everything in my power to make sure Pappy doesn't have those issues. Or even better, if you want Burp to do something that it doesn't already, let me know so that I can ~~use it to stomp them into the dust~~ improve my project.
If you're brave and want to try and contribute code, please let me know. Right now the codebase is kind of rough and I have refactored it a few times already, but I would be more than happy to find a stable part of the codebase that you can contribute to.
Another option is to try writing a plugin. It might be a bit easier than contributing code and plugins are extremely easy to integrate as a core feature. So you can also contribute by writing a plugin and letting me know about it. You can find out more by looking at [the official plugin docs](https://roglew.github.io/pappy-proxy/pappyplugins.html).
Pappy supports OS X and Linux (sorry Windows). Installation requires `pip` or some other command that can handle a `setup.py` with requirements. Once the requirements are installed, you can check that it installed correctly by running `pappy -l` to start the proxy.
Pappy projects take up an entire directory. Any generated scripts, exported responses, plugin data, etc. will be placed in the current directory so it's good to give your project a directory of its own. To start a project, do something like:
And that's it! The proxy will by default be running on port 8000 and bound to localhost (to keep the hackers out). You can modify the port/interface in `config.json`. You can list all your intercepted requests with `ls`, view a full request with `vfq <reqid>` or view a full response with `vfs <reqid>`. Right now, the only command to delete requests is `filter_prune` which deletes all the requests that aren't in the current context (look at the sections on the context/filter strings for more information on that).
If you don't want to dirty up a directory, you can run Pappy in "lite" mode. Pappy will use the default configuration settings and will create a temporary data file in `/tmp` to use. When you quit, the file will be deleted. If you want to run Pappy in lite mode, run Pappy with either `-l` or `--lite`.
In order for Pappy to view data sent using HTTPS, you need to add a generated CA cert (`certificate.crt`) to your browser. Certificates are generated using the `gencerts` command and are by default stored in `~/.pappy/certs`. This allows Pappy to act as a CA and sign any HTTPS certificate it wants without the browser complaining. This allows Pappy to decrypt and modify HTTPS requests. The certificate installation instructions are different for each browser.
You can add the CA cert to Firefox by going to `Preferences -> Advanced -> View Certificates -> Authorities -> Import` and selecting the `certificate.crt` file in the `certs` directory.
### Chrome
You can add the CA cert to Chrome by going to `Settings -> Show advanced settings -> HTTPS/SSL -> Manage Certificates -> Authorities -> Import` and selecting the `certificate.crt` file in the `certs` directory.
### Safari
For Safari (on macs, obviously), you need to add the CA cert to your system keychain. You can do this by double clicking on the CA cert and following the prompts.
I didn't search too hard for instructions on this (since Pappy doesn't support windows) and I don't own a Windows machine to try this, so if you have trouble, I'm not the one to ask. According to Google you can double-click the cert to install it to the system, or you can do `Tools -> Content -> Certificates -> Trusted Root Certificates -> Import`.
Configuration for each project is done in the `config.json` file. The file is a JSON-formatted dictionary that contains settings for the proxy. The following fields can be used to configure the proxy:
| `data_file` | The file where requests and images will be stored |
| `debug_dir` (optional) | Where connection debug info should be stored. If not present, debug info is not saved to a file. |
| `cert_dir` | Where the CA cert and the private key for the CA cert are stored |
| `proxy_listeners` | A list of dicts which describe which ports the proxy will listen on. Each item is a dict with "port" and "interface" values which determine which port and interface to listen on. For example, if port=8000 and the interface is 127.0.0.1, the proxy will only accept connections from localhost on port 8000. To accept connections from anywhere, set the interface to 0.0.0.0. |
In order to intercept and modify requests to sites that use HTTPS, you have to generate and install CA certs to your browser. You can do this by running the `gencerts` command in Pappy. By default, certs are stored `~/.pappy/certs`. This is also the default location that Pappy will look for certificates (unless you specify otherwise in `config.json`.) In addition, you can give the `gencerts` command an argument to have it put the generated certs in a different directory.
| `gencerts [/path/to/put/certs/in]` | Generate a CA cert that can be added to your browser to let Pappy decrypt HTTPS traffic. Also generates the private key for that cert in the same directory. If no path is given, the certs will be placed in the default certificate location. Overwrites any existing certs. |
| `ls [a|<num>`]| list, ls |List requests that are in the current context (see Context section). Has information like the host, target path, and status code. With no arguments, it will print the 25 most recent requests in the current context. If you pass 'a' or 'all' as an argument, it will print all the requests in the current context. If you pass a number "n" as an argument, it will print the n most recent requests in the current context. |
| `sm` | sm, site_map | Print a tree showing the site map. It will display all requests in the current context that did not have a 404 response. |
| `viq <id(s)>` | view_request_info, viq | View additional information about requests. Includes the target port, if SSL was used, applied tags, and other information. |
| `vfq <id(s)>` | view_full_request, vfq | [V]iew [F]ull Re[Q]uest, prints the full request including headers and data. |
| `vhq <id(s)>` | view_request_headers, vhq | [V]iew [H]eaders of a Re[Q]uest. Prints just the headers of a request. |
| `vfs <id(s)>` | view_full_response, vfs |[V]iew [F]ull Re[S]ponse, prints the full response associated with a request including headers and data. |
| `vhs <id(s)>` | view_response_headers, vhs | [V]iew [H]eaders of a Re[S]ponse. Prints just the headers of a response associated with a request. |
| ID | The request ID of that request. Used to identify the request for other commands. |
| Method | The method(/http verb) for the request |
| Host | The host that the request was sent to |
| Path | The path of the request |
| S-Code | The status code of the response |
| Req Len | The length of the data submitted |
| Rsp Len | The length of the data returned in the response |
| Time | The time in seconds it took to complete the request |
| Mngl | If the request or response were mangled with the interceptor. If the request was mangled, the column will show 'q'. If the response was mangled, the column will show 's'. If both were mangled, it will show 'q/s'. |
You can apply tags to a request and use filters to view specific tags. The following commands can be used to apply and remove tags to requests:
| Command | Aliases | Description |
|:--------|:--------|:------------|
| `tag <tag> [id(s)]` | tag | Apply a tag to the given requests. If no IDs are given, the tag will be applied to all in-context requests. |
| `untag <tag> [id(s)]` | untag | Remove a tag from the given ids. If no IDs are given, the tag is removed from every in-context request. |
| `clrtag <id(s)>` | clrtag | Removes all tags from the given ids. |
Request IDs
-----------
Request IDs are how you identify a request and every command that involves specifying a request will take one or more request IDs. You can see it when you run `ls`. In addition, you can prepend an ID with prefixes to get requests or responses associated with the request (for example if you modified the request or its response with the interceptor, you can get the unmangled versions.) Here are the valid prefixes:
| Prefix | Description |
|:-------|:------------|
| `u` | If the request was mangled, prefixing the ID with `u` will result in the unmangled version of the request. The resulting request will not have an associated response because it was never submitted to the server. |
| `s` | If the response was mangled, prefixing the request ID `s` will result in the same request but its associated response will be the unmangled version. |
I know it sounds kind of unintuitive. Here are some example commands that will hopefully make things clearer. Suppose request 1 had its request mangled, and request 2 had its response mangled.
*`vfq 1` Prints the mangled version of request 1
*`vfq u1` Prints the unmangled version of request 1
*`rp u1` Open the repeater with the unmangled version of request 1
*`vfs u1` Throws an error because the unmangled version was never submitted
*`vfs s1` Throws an error because the response for request 1 was never mangled
*`vfs 2` Prints the mangled response of request 2
*`vfs s2` Prints the unmangled response of request 2
*`vfq u2` Throws an error because request 2's request was never mangled
*`vfs u2` Throws an error because request 2's request was never mangled
### Passing Multiple Request IDs to a Command
Some arguments can take multiple IDs for an argument. To pass multiple IDs to a command, separate the IDs with commas **(no spaces!)**. A few examples:
*`viq 1,2,u3` View information about requests 1, 2, and the unmangled version of 3
*`gma foo 4,5,6` Generate a macro with definitions for requests 4, 5, and 6
The context is a set of filters that define which requests are considered "active". Only requests in the current context are displayed with `ls`. By default, the context includes every single request that passes through the proxy. You can limit down the current context by applying filters. Filters apply rules such as "the response code must equal 500" or "the host must contain google.com". Once you apply one or more filters, only requests/responses which pass every active filter will be a part of the current context.
| `f <filter string>` | filter, fl, f |Add a filter that limits which requests are included in the current context. See the Filter String section for how to create a filter string |
Where `<field>` is some part of the request/response, `<comparer>` is some comparison to `<value>`. For example, if you wanted a filter that only matches requests to `target.org`, you could use the following filter string:
Also **if you prefix a comparer with 'n' it turns it into a negation.** Using the previous example, the following will match any request except for ones where the host contains `target.org`:
For fields that are a list of key/value pairs (headers, get params, post params, and cookies) you can use the following format:
```
<field><comparer1><value1>[ <comparer2><value2>]
```
This is a little more complicated. If you don't give comparer2/value2, the filter will pass any pair where the key or the value matches comparer1 and value1. If you do give comparer2/value2, the key must match comparer1/value1 and the value must match comparer2/value2 For example:
```
Filter A:
cookie contains Session
Filter B:
cookie contains Session contains 456
Filter C:
cookie ncontains Ultra
Cookie: SuperSession=abc123
Matches A and C but not B
Cookie: UltraSession=abc123456
Matches both A and B but not C
```
### List of fields
| Field Name | Aliases | Description | Format |
|:--------|:------------|:-----|:------|
| all | all | The entire request represented as one string | String |
Scope is a set of rules to define whether Pappy should mess with a request. You define the scope by setting the context to what you want the scope to be and running `scope_save`. The scope is saved in the data file and is automatically restored when using the same project directory.
Any requests which don't match all the filters in the scope will be passed straight to the browser and will not be caught by the interceptor or recorded in the data file. This is useful to make sure you don't accidentally do something like log in to your email through the proxy and have your plaintext username/password stored.
| `scope_save` |`scope_save`| Set the current context to be the scope |
| `sr` |`scope_reset`, `sr`| Set the current context to the scope |
| `scope_delete` |`scope_delete`| Clear the scope (everything's in scope!) |
| `scope_list` |`scope_list`, `sls`| List all the filters that are applied to the scope |
### Built-In Filters
Pappy also includes some built in filters that you can apply. These are things that you may want to filter by but may be too tedius to type out. The `fbi` command also supports tab completion.
This feature is like Burp's proxy with "Intercept Mode" turned on, except it's not turned on unless you explicitly turn it on. When the proxy gets a request while in intercept mode, it lets you edit it before forwarding it to the server. In addition, it can stop responses from the server and let you edit them before they get forwarded to the browser. When you run the command, you can pass `req` and/or `rsp` as arguments to say whether you would like to intercept requests and/or responses. Only in-scope requests/responses will be intercepted (see Scope section).
The interceptor will use your EDITOR variable to decide which editor to edit the request/response with. If no editor variable is set, it will default to `vi`.
| `ic <req,rsp>+` | `intercept`, `ic` | Begins interception mode. Press enter to leave interception mode and return to the command prompt. Pass in `request` to intercept requests, `response` to intercept responses, or both to intercept both. |
This feature is like Burp's repeater (yes, really). You choose a request and Pappy will open vim in a split window with your request on the left and the original response on the right. You can make changes to the request and then run ":RepeaterSubmitBuffer" to submit the modified request. The response will be displayed on the right. This command is bound to `<leader>f` by default, but you can bind it to something else too in your vimrc (I think, dunno if vim will complain if the function undefined which it will be for regular files). This command will submit whatever buffer your cursor is in, so make sure it's in the request buffer.
Macros are Pappy's version of Burp's intruder. You can use macros to make automated requests through the proxy and save them to the data file. A macro file is any python script file in the current directory that is in the form `macro_<name>.py`. An example project directory with macros would be:
```
$ ls -l
-rw-r--r-- 1 scaryhacker wheel 150 Nov 26 11:17 config.json
-rw------- 1 scaryhacker wheel 2639872 Nov 26 17:18 data.db
-rw-r--r-- 1 scaryhacker wheel 471 Nov 26 18:42 macro_blank.py
-rw-r--r-- 1 scaryhacker wheel 264 Nov 26 18:49 macro_hackthensa.py
-rw-r--r-- 1 scaryhacker wheel 1261 Nov 26 18:37 macro_testgen.py
-rw-r--r-- 1 scaryhacker wheel 241 Nov 26 17:18 macro_test.py
```
In this case we have a `blank`, `hackthensa`, `testgen`, and `test` macro. A macro script is any python script that defines a `run_macro(args)` function and a `MACRO_NAME` variable. For example, a simple macro would be:
You can place this macro in your project directory then load and run it from Pappy. When a macro is run, arguments are passed from the command line. Arguments are separated the same way as they are on the command line, so if you want to use spaces in your argument, you have to put quotes around it.
You'll need to run `lma` every time you make a change to the macro in order to reload it. In addition, any code outside of the `run_macro` function will be run when it the macro gets loaded.
### Generating Macros From Requests
You can also generate macros that have Pappy `Request` objects created with the same information as requests you've already made. For example:
If you enter in a value for `SHORT_NAME`, you can use it as a shortcut to run that macro. So if in a macro you set `SHORT_NAME='tm'` you can run it by running `pappy> rma tm`.
| `lma [dir]` | `load_macros`, `lma` | Load macros from a directory. If `dir` is not given, use the current directory (the project directory) |
| `rma <macro name>` | `run_macro`, `rma` | Run a macro with the given name. You can use the shortname, filename, or long name. |
| `gma <name> [id(s)]` | `generate_macro`, `gma` | Generate a macro with the given name. If request IDs are given, the macro will contain request objects that contain each request. |
| `rpy <id(s)>` | `rpy` | Print the Python object definitions for each of the given ids |
The main method of interacting with the proxy is through `Request` objects. You can submit a request with `req.sumbit()` and save it to the data file with `req.save()`. The objects also have attributes which can be used to modify the request in a high-level way. You can see the [full documentation](https://roglew.github.io/pappy-proxy/pappyproxy.html#module-pappyproxy.http) for more details on using these objects.
Dict-like objects are represented with a custom class called a `RepeatableDict`. Again, look at the docs for details. For the most part, you can interact with it like a normal dictionary, but don't be surprised if it's missing some methods you would expect.
| port | Yes | Integer | The port the request is/was sent to |
| post_params | Yes | RepeatableDict | Post parameters |
| raw_data | Yes | String | The data part of the request |
| raw_headers | No | String | The text of the headers section of the request |
| reqid | Yes | Integer | The ID of the request. If set when save() is called, it replaces the request with the same id in the database |
| response | Yes | Response | The associated response for the request |
| rsptime | No | Datetime Delta | The time it took to complete the request. Set when submit() is called |
| status_line | Yes | String | The status line of the request (ie 'GET / HTTP/1.1') |
| time_end | Yes | Datetime | The time when the request was completed |
| time_start | Yes | Datetime | The time when the request was started |
| unmangled | Yes | Request | If the request was mangled, the unmangled version of the request |
| url | Yes | String | The URL of the request (ie 'https://www.google.com') |
| url_params | Yes | RepeatableDict | The URL parameters of the request |
| verb | Yes | String | The verb used for the request (ie GET, POST, PATCH, HEAD, etc). Doesn't have to be a valid verb. |
| version | Yes | String | The version part of the status line (ie 'HTTP/1.1') |
Request methods:
| Function | Description |
|:--|:--|
| submit() | Submit the request through the proxy. Does not save the request to the data file |
| save() | Save the request, its unmangled version, its associated response, and the unmangled version of the response to the database |
And here is a quick list of attributes that you can use with `Response` objects:
| Attribute | Settable? | Data Type | Description |
|:--|:--|:--|:--|
| cookies | Yes | RepeatableDict | Cookies set by the response |
| headers | Yes | RepeatableDict | The headers of the response |
| response_code | Yes | Integer | The response code of the response |
| response_text | Yes | String | The text associated with the response code (ie OK, NOT FOUND)
| rspid | Yes | Integer | The response id of the response. If this is the same as another response in the database, calling save() on the associated request will replace that response in the database |
| unmangled | Yes | Response | If the response was mangled, this will refer to the unmangled version of the response. Otherwise it is None |
| version | Yes | String | The version part of the status line of the response (ie 'HTTP/1.1') |
| raw_headers | No | String | A text version of the headers of the response |
| status_line | Yes | String | The status line of the response |
| raw_data | Yes | String | The data portion of the response |
| full_response | No | String | The full text version of the response including headers and data |
Like I said, these interfaces are prone to change and will probably crash when you use them. If you get a traceback, send me an email so I can fix it.
There are also a few functions which could be useful for creating requests in macros. It's worth pointing out that `request_by_id` is useful for passing request objects as arguments. For example, here is a macro that lets you resubmit a request with the Google Bot user agent:
```
## macro_googlebot.py
from pappyproxy.http import Request, get_request, post_request, request_by_id
| get_request(url, url_params={}) | Returns a Request object that contains a GET request to the given url with the given url params |
| post_request(url, post_params={}, url_params={}) | Returns a Request object that contains a POST request to the given url with the given url and post params |
Intercepting macros let you mangle requests as they pass through the proxy. Similarly to normal macros, an intercepting macro is any python script with an "int" prefix. For example, `int_name.py` would be a valid intercepting macro name. They are also loaded with the `lma` command. An intercepting macro can define two functions: `mangle_request` or `mangle_response`. Both requests only take a `Request` object as a parameter. `mangle_request` returns either a new, modified Request object to change it, or it can return the original object to not mangle it. The `mange_response` must return a `Response` (not request!) object. The request passed in to `mangle_response` will have an associated response with it. If you want to modify the response, copy `request.response`, make modifications, then return it. If you would like to pass it through untouched, just return `request.response`.
Note, that due to twisted funkyness, *you cannot save requests from intercepting macros*. Technically you **can**, but to do that you'll have to define `async_mangle_request` (or response) instead of `mangle_request` (or response) then use `Request.async_deep_save` which generates a deferred, then generate a deferred from `async_mangle_requests` (inline callbacks work too). If you've never used twisted before, please don't try. Twisted is hard. Plus the mangled request will be saved before it is submitted anyways.
Confusing? Here are some example intercepting macros:
In addition, you can use an `init(args)` function to get arguments from the command line. If no arguments are passed, args will be an empty list. Here is an example macro that does a search and replace:
You can use this macro to do any search and replace that you want. For example, if you wanted to replace "Google" with "Skynet", you can run the macro like this:
You can use the following commands to start/stop intercepting macros
| Command | Aliases | Description |
|:--------|:--------|:------------|
| `lma [dir]` | `load_macros`, `lma` | Load macros from a directory. If `dir` is not given, use the current directory (the project directory) |
| `rim <macro name>` | `run_int_macro`, `rim` | Run an intercepting macro. Similarly to normal macros you can use the name, short name, or file name of the macro. |
| `sim <macro name> [args]` | `stop_int_macro`, `sim` | Stop an intercepting macro. If arguments are given, they will be passed to the macro's `init(args)` function if it exists. |
| `lim` | `list_int_macros`, `lsim` | List all enabled/disabled intercepting macros |
You can watch in real-time what requests are going through the proxy. Verbosisty defaults to 1 which just states when connections are made/lost and some information on what is happening. If verbosity is set to 3, it includes all the data which is sent through the proxy and processed. It will print the raw response from the server, what it decodes it to, etc. Even if you don't run this command, all the information is stored in the dubug directory (the directory is cleared every start though!)
| `log [verbosity]` | View the log at the given verbosity. Default verbosity is 1 which just shows connections being made/lost and some other info, verbosity 3 shows full requests/responses as they pass through and are processed by the proxy |
This is a list of other random stuff you can do that isn't categorized under anything else. These are mostly commands that I found that I needed while doing a test and just added. They likely don't do a ton of error checking.
| Command | Aliases | Description |
|:--------|:--------|:------------|
| `dump_response <reqid> [filename]` | `dump_response` | Dumps the data from the response to the given filename (useful for images, .swf, etc). If no filename is given, it uses the name given in the path. |
| `export <req|rsp> <reqid>` | `export` | Writes either the full request or response to a file in the current directory. |
If you don't have any intercepting macros running, Pappy will forward data to the browser as it gets it. However, if you're trying to mangle messages/responses, Pappy will need to download the entire message first.
Plugins
-------
Note that this section is a very quick overview of plugins. For a full description of how to write them, please see [the official docs](https://roglew.github.io/pappy-proxy/pappyplugins.html).
It is also possible to write plugins which are reusable across projects. Plugins are simply Python scripts located in `~/.pappy/plugins`. Plugins are able to create new console commands and maintain state throughout a Pappy session. They can access the same API as macros, but the plugin system is designed to allow you to create general purpose commands as compared to macros which are meant to be project-specific scripts. Still, it may not be a bad idea to try building a macro to do something in a quick and dirty way before writing a plugin since plugins are more complicated to write.
A simple hello world plugin could be something like:
```
## hello.py
import shlex
def hello_world(line):
if line:
args = shlex.split(line)
print 'Hello, %s!' % (', '.join(args))
else:
print "Hello, world!"
###############
## Plugin hooks
def load_cmds(cmd):
cmd.set_cmds({
'hello': (hello_world, None),
})
cmd.add_aliases([
('hello', 'hlo'),
('hello', 'ho'),
])
```
You can also create commands which support autocomplete:
return [n for n in _AUTOCOMPLETE_NAMES if n.startswith(text)]
###############
## Plugin hooks
def load_cmds(cmd):
cmd.set_cmds({
'hello': (hello_world, complete_hello_world),
})
cmd.add_aliases([
('hello', 'hlo'),
])
```
Then when you run Pappy you can use the ``hello`` command:
```
$ pappy -l
Temporary datafile is /tmp/tmpBOXyJ3
Proxy is listening on port 8000
pappy> ho
Hello, world!
pappy> ho foo bar baz
Hello, foo, bar, baz!
pappy> ho foo bar "baz lihtyur"
Hello, foo, bar, baz lihtyur!
pappy>
```
### Should I Write a Plugin or a Macro?
A lot of the time, you can get away with writing a macro. However, you may consider writing a plugin if:
* You find yourself copying one macro to multiple projects
* You want to write a general tool that can be applied to any website
* You need to maintain state during the Pappy session
My guess is that if you need one quick thing for a project, you're better off writing a macro first and seeing if you end up using it in future projects. Then if you find yourself needing it a lot, write a plugin for it. You may also consider keeping a `mine.py` plugin where you can write out commands that you use regularly but may not be worth creating a dedicated plugin for.
There are some settings that apply to Pappy as a whole and are stored in `~/.pappy/global_config.json`. These settings are generally for tuning performance or modifying behavior on a system-wide level. No information about projects is put in here since it is world readable. You can technically add settings in here for plugins that you write, but if it's at all possible, please keep settings in the normal project config.
Settings included in `~/.pappy/global_config.json`:
| Setting | Description |
|:--------|:------------|
| cache_size | The number of requests from history that will be included in memory at any given time. Set to -1 to keep everything in memory. See the request cache section for more info. |
### I still like Burp, but Pappy looks interesting, can I use both?
Yes! If you don't want to go completely over to Pappy yet, you can configure Burp to use Pappy as an upstream proxy server. That way, traffic will go through both Burp and Pappy and you can use whichever you want to do your testing.
How to have Burp forward traffic through Pappy:
1. Open Burp
2. Go to `Options -> Connections -> Upstream Proxy Servers`
3. Click `Add`
4. Leave `Destination Host` blank, but put `127.0.0.1` in `Proxy Host` and `8000` into `Port` (assuming you're using the default listener)
5. Configure your browser to use Burp as a proxy
### Why does my request have an id of `--`?!?!
You can't do anything with a request/response until it is decoded and saved to disk. In between the time when a request is decoded and when it's saved to disk, it will have an ID of `--`. So just wait a little bit and it will get an ID you can use.
I do some stuff to try and keep speed and memory usage to reasonable levels. Unfortunately, things might seem slow in some areas. This is where I try and explain why those exist. Honestly, you probably don't care about this, but I'd rather have it written down and have nobody read it than just leave people in the dark.
### Request Cache / Memory usage
For performance reasons, Pappy by default will not store every request in memory. The cache will store a certain number of the most recently accessed requests in memory. This means that if you go through all of history, it could be slow (for example running `ls a` or `sm`). If you have enough RAM to keep everything in memory, you can set the request cache size to -1 to just keep everything in memory. However, even if the cache size is unlimited, it still won't load a request into memory untill you access it. So if you want to load everything in memory, run `ls a`.
By default, Pappy will cache 2000 requests. This is kind of heavy, but it's assumed you're doing testing on a reasonably specced laptop. Personally, I live on the edge and use -1 until I run into memory issues.
