Pickaxe User Manual Guide

This is our canonical guide to Pickaxe. If you're getting started with Pickaxe, we recommend reading it. If you're looking for a quicker overview of the product, you can check out this 10-min video tour of Pickaxe.

Pickaxe is a no-code, self-serve platform for creating AI tools, sharing them with users, and monetizing them with paywalls and usage limits. If you want to build AI tools for yourself, your company, your clients, or your customers, Pickaxe is a simple place to do that. We'll cover every part of that process in this guide.

You can also ask our Pickaxe Doorman any questions. He is an AI chatbot trained on this guide as well as many other Pickaxe materials. Talking with the Doorman might be faster and simpler than reading static text. You can also check out our Frequently Asked Questions page. Additionally, if you just want to hire someone to build your Pickaxes for you, you can always hire a Pickaxe Expert. These experts are customers who have mastered the platform and are now frequently hired by other customers to build and manage AI tools on Pickaxe.

Let's go through Pickaxe, piece by piece.

‍

How to build a Pickaxe

Forms vs Chatbots

A Pickaxe is an AI tool built on top of an LLM (like OpenAI's GPT-4 or Anthropic's Claude Sonnet). In addition to the underlying AI model's out-of-box capability, you can write special instructions in the form of a prompt, give it special knowledge by uploading your own data, and even connect it to other softwares and APIs via Actions.

You can build two different types of Pickaxes. There are Forms. And there are Chatbots. The principles are very similar but the end-user facing experience is a little different.

A Chatbot is a pretty self-explanatory. It is a conversation box where the end-user types whatever they want. When the user sends a message, the Chatbot responds. It is an open-ended, back and forth conversation with no definite end. Much like ChatGPT.

A Form is a series of input fields. These can include text input fields, multiple choice, file upload fields, etc. Once a user hits the “submit” button, the prompt will be run using the inputs provided by the end-user. Then the output will be displayed to the end-user. Forms can even be configured so they output automatically opens up in a chat. Chatbots are the most popular type of AI tool. However, some customers prefer Forms because it is more structured than a Chatbot.

The difference between a Chatbot and a Form mainly exist in the experience of the end-user using the tool. The experience of building a Form or a Chatbot is very similar.

The Builder

The Pickaxe Builder is a two-panel interface. In the left-hand side you build the tool (write the prompt, add files, etc.) On the right-hand side you can test your tool live at any moment. The combination of the two lets you rapidly build and test your tool in a quick iterative loop. You can make a change in the left-hand panel and instantly test its effect in the right-hand panel.

The builder allows you write a prompt (the instructions for how your tool should behave), add files to the Knowledge Base (such as PDFs, word docs, spreadsheets, websites, PowerPoints, even YouTube videos) that your tool will draw from to inform its answers, customize some superficial design aspects (icon, title, placeholder text) and integrate external APIs via our Actions system (which allows your tool to call and use other pieces of software).

In the Prompt tab you can write your prompt, or the instructions for how your tool should behave. Prompting is the practice of writing instructions for an AI model in normal language. It's often called "Prompt engineering" or "Prompt Design". It’s about using the right words, format, and structure to guide the AI’s output. There's a lot of advice out there on writing effective prompts. The best advice is to write clearly and explicitly about what you want the AI to do.

In the Model Dropdown you can select which AI model you want to power your tool. It's located right underneath the box where you write your prompt. Don't stress too much about which model to use. The default selection of GPT-4o mini is a smart, fast, cheap, and capable model. You should only change the model if you have a specific reason or want to experiment. For example, many copywriters who care about prose style often claim Claude is a preferable model.

In the Configure tab, you can adjust some settings for your tool, customize the icon + image, and do some other miscellaneous things. The default settings are good for most standard use cases, but you may want to come here to change a couple advanced things like token settings. If you want to change the profile image of your Pickaxe, change the placeholder text in the chat input area, or add a special chat icon, you can do all that here.

In the Knowledge tab, you can add files to your Knowledge Base. The Knowledge Base is a RAG (Retrieval Augmented Generation) system. If you don’t know what that is, don’t worry. There’s a section about that later on. Basically you upload important pieces of knowledge and your tool will use those bits of knowledge to better inform its responses.

In the Actions tab you can connect create Actions. The Actions system is a way to let your tool to commit actions a language model wouldn't normally be able to do. This might be generating an image, triggering a Zapier webhook, or sending an email. There's another section about that later.

Here’s a video tutorial about the Chatbot builder.

Here’s a video tutorial about the Form builder.

Pickaxes with image generation or file upload

There are some basic added capabilities that are very popular with Pickaxes. These include the ability for a Pickaxe to generate an image and the ability for end-users to upload files into the Pickaxe. Both of these capabilities can be turned on with a simple toggle.

End-user Upload: If you want to allow your end-users to upload files into your Pickaxe, simply toggle the "allow users to upload files" setting in the prompt tab of the Builder. This will add a little file upload paperclip icon to your Pickaxe. End-users will be able to upload text files (word docs, PDFs, txt files, spreadsheets, etc.) into your Pickaxe. These end-user uploads only last for a single session-- they do not persist across sessions with the user. So if an end-user uploads their resume into your Pickaxe chatbot, then refreshes a new conversation, the Pickaxe will not remember the resume. Essentially, when an end-user uploads a file, we attempt to dump the entire contents into the conversation as if the end-user typed and sent it. If the uploaded file is small enough to fit within the Max Input Length (a setting inside the Configure tab), the entire document will get used directly in the conversation. If it’s too big, it gets turned into vector embeddings and accessed with a retrieval system (RAG) throughout the conversation. Here's a help post that explains in more detail how end-user document upload works.

Image recognition: If you want your Pickaxe to recognize images, you'll want to make sure you've enabled end-user upload and that you are using an OpenAI model like GPT-4o. Only the OpenAI models support image recognition. This means if an end-user uploads an image into your Pickaxe, it will be able to understand what the image is. Here's a video tutorial about building a Pickaxe with image recognition.

Image Generation: If you want to allow your Pickaxe to generate images, simply toggle on the "use default image generation" toggle. This will allow your Pickaxe to use the AI image model Flux to generate image. It will only generate images when it feels appropriate. For example, if you enable image generation on a copywriter chatbot, it probably won't generate images in most of its outputs unless the end-user specifically states or implies they want an image. You can always add instructions to your chatbot about the proper conditions of when and when not to generate images. It's pretty smart at following these directions! You can connect other image models such as DALL-E within the Act tab of the builder.

‍

The Knowledge Base

How does the Knowledge Base work?

The Knowledge Base in Pickaxe is a system that lets you upload documents that would normally be too large for an AI model to read. This information is used by your Pickaxe to better inform its answers. The Knowledge Base system is a RAG (Retrieval Augmented Generation) system. When you upload a file, our system breaks it into small uniform chunks, turns those into vector embeddings, and stores them for fast semantic search. Importantly, your Pickaxe does not read or memorize the entire document every time it answers a question. Instead, when an end-user sends a message, the system scans all the stored chunks, scores them based on relevance, and selects only the most relevant ones—usually just a few paragraphs. These chunks are then inserted into the chatbot’s context so it can answer accurately and efficiently. In other words, you can upload millions of words into the Knowledge Base, and your Pickaxe will look at the most relevant few thousand each time it answers a question.

Uploading Files

You can upload files into the Knowledge Base under the Knowledge tab in the Builder. Whenever you upload a file into the Knowledge Base, it grabs all the text in the file and adds it to the Knowledge Base. That means if you add a PDF with several images in it, the images will not be read. Only text is added.

You can add all sorts of files (txt files, word docs, PDFs, powerpoints, spreadsheets, etc.) to the Knowledge Base. In addition to text documents, you can also add webpages and Youtube videos. If you add a youtube video to the Knowledge Base, the system will grab the audio transcript of the Youtube video (everything that is said aloud in the video) and add that to the Knowledge Base. If you add a webpage, a third-part web-scraping service will scrape all the text on that webpage and add it to the Knowledge Base. Not all webpages cooperate equally well with the web-scraper.

When a file is added, it is automatically scraped for text, then chunked into uniform chunks. You can always click on a file in the Knowledge Base to see exactly what's being pulled from the file. Clicking on the file opens up a Chunk Explorer view where you can view every chunk.

The chunking process sometimes orphans information (splits it between two chunks). It's just how the system has to work. However, the system handles spreadsheets in a special way. When you upload a spreadsheet, Pickaxe uses a special system that converts each row into a chunk, using the header row as properties. This allows the rows in the spreadsheet to stay in tact to be read within context of their headers. Here's a community post that explains in more detail how spreadsheets are added to the Knowledge Base.

Each Studio has its own Knowledge Base. On a Pickaxe-by-Pickaxe basis, you can specify which files in the Knowledge Base you want your Pickaxe to pull from.

The Knowledge Base has limits on the total amount of knowledge you can add. Free account can upload 3 files per Studio. Gold accounts can upload 50 files per Studio. And Pro accounts can upload unlimited files per Studio.

Knowledge Base Settings

You can control how the Knowledge Base pulls information by adjusting settings under the Knowledge tab. There are two controls for the Knowledge Base: Relevance Cutoff & Amount. By playing around with these two settings you can influence how your Pickaxe draws information from the Knowledge Base.

The Relevance Cutoff is basically a filter for how semantically relevant a chunk needs to be in order to pulled. The relevance threshold controls how closely related a chunk must be to the user’s message to be included in the Pickaxe's context. Setting this to a higher number (like 0.9) means it will only return highly relevant chunks. Setting it to a lower number (like 0.4) will mean many more chunks are referenced. We recommend playing around with these settings.

The Amount settings is a limit on how many chunks total can be referenced. This controls how much text from the Knowledge Base can be added to the chatbot’s context at once. Raising the limit gives the chatbot access to more information, but too much text can sometimes dilute the quality of its response.

The Chunk Explorer is a view we provide to builders so they can see how their Knowledge Base is working.Within the Builder, you can click on a file in the Knowledge Base to open it up in the Chunk Explorer. This is a view where you can see all the chunks. You can even enter sample queries in the search bar and see which chunks come up and their corresponding relevance score. For any response generated in the Builder's test screen (the right-hand panel) you can click on the magnifying glass icon underneath the answer to see which chunks it pulled up. This magnifying glass is called the Message Insights.

Finally, we provide the ability to add some Context to your Knowledge Base. The default setting there is fine. Otherwise, it's a way to provide extra context to the information you're putting into the Knowledge Base. For example, if you're uploading highly formatted JSON about products in a warehouse, you might use the context area to explain some of the shorthand that might not be apparent. Usually, we recommend leaving this alone.

Here’s a video tutorial about how to setup Knowledge Bases.

Connecting APIs to your Pickaxe

Actions

Connecting an Action is a way to let your Pickaxe do things beyond just generating text. Normally, a Pickaxe can read and write text, but it can’t actually do anything—like send an email, post to Slack, or generate a PDF. That’s where Actions come in. An Action is like giving your Pickaxe a button it can press to trigger an outside service. You write natural-language instructions (a trigger prompt) that tell the Pickaxe when and how to press that button. Your Pickaxe still runs off a no-code prompt, but now it can act on what it knows—by calling external tools, passing info out, or pulling results back in. It’s prompt engineering extended to real-world software.

Actions in Pickaxe let your AI tools connect to other software like Slack, Zapier, or email. They allow your Pickaxe to do things like send emails, create images, or push data to other apps—things language models can’t do on their own.

To connect an Action, go into the Actions tab and select an existing Action you want to connect. When you click on the Action it will ask you to fill out a few setup fields. These configurations are different across Actions depending on what it does. A common field among all Actions is the trigger prompt. The trigger prompt tells your Pickaxe when and how to use the Action. It's a prompt, so it's written in plain language.

You can also create your own Action. This is for more advanced users. It allows you to essentially write your own Action in Python that you can teach your Pickaxe to call and use.

Pickaxe users on a free plan can connect 1 Action per tool. Users on the Gold plan can connect 3 Actions per tool. And users on the Pro plan can connect unlimited Actions per tool.

Here’s a video tutorial about setting up Actions.

Here’s a video tutorial about creating your own advanced Actions.

An easy shortcut in Actions is to just connect an Action for an automation workflow software like Zapier, Make, Pabbly, or IFTTT. This softwares are great places to build advanced workflows. And then you can connect a single Action to trigger the automation. Here's a video tutorial about creating a Make webhook Action to send and receive information.

Actions are a really cool part of Pickaxe. They can also seem kind of intimidating at first. If you have questions, you can go to the Actions section of the community forum.

‍

Studios

What is a Studio?

Studios are workspaces that contain your Pickaxes. This is where you manage, share, embed, limit, monetize, and monitor your Pickaxes. Whenever you build a Pickaxe, it must go into a Studio.

The Studio has a left-hand navigation menu with 7 tabs. We'll go over each one.

In the Pickaxes tab, you'll find a list of all your tools in your Studio. Here you can edit the tools, view information about them, and create embeds you can put into third-party apps and websites. If you click on a Pickaxe in the list, you'll see details about that specific Pickaxe. You can also use that Pickaxe here too.

‍

In the Design tab, you can customize the visual look of your Pickaxe. These customizations include colors, fonts, text sizing, logos, white-labeling, and more. When you enter the Design tab, you'll see a live preview of your tools so you can quickly make changes.

In the Users tab, you'll see all the users of your Studio and their activity. In the Activity section you can see what users are doing and how your tool is performing. If you're testing a tool, it's important to monitor activity to make sure it's performing as desired. In the Manage Users section, you can see all the Users of your tool. This will include accounts with email addresses and users who have not created an account (tracked by IP address). By clicking on a User you can see more information about them and even give them free credits or upgrades, and invite new Users. You can invite new Users here too. In the User Memories section, you can also create and manage User Memories which are unique traits your Studio remembers on a user-by-user basis in order to provide a more customized experience.

In the Knowledge Base tab, you can manage the files in your Studios Knowledge Base.

In the Access tab, you can control restrictions on access to your Studio and the tools inside. These restrictions include who can you use your Studio (public vs invite-only), usage limits, and paywalls. We'll discuss this topic at length later in the guide.

In the Settings tab, you'll find miscellaneous settings for your entire Studio such as the name, logo, whether it's hosted at a custom domain, whether it uses Pickaxe credits or your own API key, and more. There is also a section to insert special code into the header or footer of your studio (for ad-tracking, analytics, etc.).

At the very bottom is the Website tab. This is a special tab. Each Studio has a basic website attached to it where end-users can see all the Pickaxes in one convenient place. We'll talk more about that next.

‍

Studio Websites

The Studio Websites are an easy way to host and share all your tools in an elegant, configurable container for other people to use. The website lists all your tools in a left-hand navigation. Clicking on a tool opens it up. Studios also come with basic functionality like user authentication, payment rails, user management, and usage limiting.

Listed vs Unlisted Pickaxes

Remember your Studio is first and foremost a workspace to build and share Pickaxes. The Website is just one way to share these tools. Tools in your Studio are either listed in your website or unlisted.

If a tool is Listed, it means the tool will appear in your Studio website. If you share a link to that tool you are in fact sharing a link to your Studio Website as a result. The user will be able to see a left-hand nav bar with other tools in the Studio Website.

If a tool is Unlisted, it means the will not appear in your Studio website. If you share a link to that tool the page will only include the tool. Unlisted tools still inherit Studio-level styling and usage restrictions. They're just one-off pages.

You can list/unlist a tool at any time from the Pickaxes tab.

Editing your Studio Website

To edit your website, click on the edit button. This will open up a Visual Website Editor where you can customize your website.

Your website's visual style inherits the styling choices you made in the Design tab. You can always go back to the Design tab to make further changes.

In a Studio Website, you can add pages by clicking the "+" button. You can add three types of things: Pickaxes, Content Pages, and Folders.

Pickaxes are your tools. A Pickaxe does not automatically appear in your Studio Website unless you specifically add it. When adding Pickaxes within the Studio Website you can bulk add them.

Content Page are simple pages where you can add text, images, and even video embeds. In a Content Page you can add blocks of content. These can be either Markdown sections or HTML sections. You can read more about Content Pages here.

Folders are buckets where you can put Pickaxes and Content Pages. Folders are collapsible and expandable. Builders like to group related tools inside of Folders.

You can re-order your pages at any time, by clicking and dragging them up and down the left-hand menu. Dragging your pages will re-order them within your Studio Website.

Whenever you want to change text on a page, make sure to click "Edit Page" in the top right of the page. And whenever you want to save your edits make sure to click "Save Edits" in the same top right corner.

‍

Public vs Invite-only Studios

There are two types of Studios, Public and Invite-Only. You can find and change this high-level setting within the Access tab of your Studio.

If a Studio is "Public" it means anyone (even strangers online) can visit your Studio. A user still needs the link to your Studio in order to access it. Making a Studio "Public" does not mean it is listed anywhere special within Pickaxe. It jsut means visitors can see what's inside. Even if a Studio is "Public", users are still subject to your Access settings. For example, if allow 0 Guest uses, visitors will not be able to use any tools unless they pay.

If a Studio is "Invite-Only" it means only users who have been specifically invited or approved by the owner may access the Studio. Any un-approved user who tries to access an Invite-Only Studio will be hit with a login wall, blocking them from seeing or accessing the Studio. Users can gain access to an Invite-Only Studio by being invited by the owner or requesting access.

Inviting a user to a Studio is the same process for both Public and Invite-Only Studios. Go to the Users tab, click on Manage Users, and then click the "+ Add User" button. Here you can invite users by email. You can even set the invite to automatically give them a certain access level. If a user requests access to a studio, the Studio owner will receive an email asking them to approve the user.

‍

Monetizing, Paywalling, & Restricting Usage

Within the Access tab of the Studio you can paywall your tools, restrict usage, and connect a Stripe account.

Credits are the system you will use to limit your users. 1 credit = 1 use. Whenever an end-user of your Studio causes one of your Pickaxes to generate a response, they are spending 1 credit.

Guest Settings are the restrictions that affect visitors to your studios who have not logged-in or not bought a product yet. You can find a Guest Settings section in the Access tab. Here you can set the number of credits Guests get. These credits refresh every 30 days. Give a think about whether you want Guests to have 0 credits, 1 credits, or 100 credits.

Products are the special bundles of credits and access you can sell to your end-users. In the Access tab, you will find a dedicated Products & Paywalls section.Here you can create Products. When you create one you will set a price, a number of credits, and a list of Pickaxes it unlocks. If you put a Pickaxe in a Product, you are paywalling it behind that Product-- the Pickaxe will only be available to users who’ve purchased that Product. Any Pickaxes that are not included in Products remain available to all users by default, including guests. Within Products, there are several other customizations such as billing frequency (one-time, monthly, yearly, etc.) and even visibility (hidden or visible).

Once you created a Product, it will appear on your Studio so your end-users may purchase it. You will also see it across the backend of your Studio for certain admin operations. For example, when you Manage Users, you can manually upgrade them to any Product.

In order to charge for Products, you'll need to connect your Stripe account. Here is a post about connecting your Stripe account to Pickaxe.

Managing Users

Within the Users tab of your Studio, you can manage users and monitor activity across your Studio.

Under the Activity tab, you can see the usage across your studio. Each conversation will have a high-level AI-generated summary. You can also click in and see the actual conversation.

Under the User Memories tab, you can set User Memories that your STudio will collect and remember about users. You can read more about the User Memory feature here.

Under the Manage Users tab, you can see all your users, check out their usage, and perform admin actions. Clicking on a user will show you a card with all their details. Here you can give them extra credits, manually assign them a Product, and more.

‍

Embedding Pickaxes

One of the most popular features in Pickaxe is embedding your tools into third-party websites. Pickaxe allows you to customize a slick iframe or script embed, white-label it, and even set basic usage restrictions.

To embed a Pickaxe, click on the Pickaxe in your Studio. This will open an overview page of that Pickaxe. Look for the "+ Create Embed" button and click it.

This will open the Embed Customization screen. On this screen, you can customize all sorts of things such as font, color, sizing, white-labeling, embed type, and more. You'll see a live preview of your embed in the middle of screen that reflects your changes in real-time. In the Embed Customization screen you can also set a unique usage limit for this embed. When someone using your embedded Pickaxe on a third-party website (not Pickaxe) hits the usage limit, they will receive a custom message that redirects them to the location of your choice.

Once you're happy with your embed, click "Save". The code to copy/paste your embed into a website is in the top of the page at all times. If you've saved an embed, you can go in and edit the stylization of it and you will not need to re-embed it.

Still have questions?

That's a basic overview of Pickaxe. There's a lot more too. If you have questions, try asking the Doorman. You can also email info@pickaxeproject.com.

Pickaxe User Manual: How to do everything in Pickaxe

Related articles

December 21, 2023

Semantic Prompt Design: A comprehensive guide

December 14, 2023

How to design an AI persona bot: Pickaxe chatbot builder walkthrough

December 15, 2023

How to use books and documents to make a chatbot: Building a Charlie Munger chatbot