Build AI agents for anything, sell them to anyone

What is Pickaxe?

Pickaxe is a code-free AI Agent platform designed for creators, consultants, organizations, and businesses, enabling tasks such as creation, deployment, user management, usage billing, and sales.

The long English titles in the database are the product slogans; the current name of the product is Pickaxe. Each AI tool on the platform is also referred to as a Pickaxe.

Main functions

• AI Agent Builder: Describe the objectives in natural language; AI Helper will suggest the appropriate settings and display the changes before they are applied, or it is possible to make adjustments manually within the editor.
• Multiple tool formats: It is possible to create chat agents, form-based tools, or combined tools that allow for continued conversation after a form has been submitted.
• Multi-model switching: It is possible to compare and switch between different model providers, without having to stick each agent to a single model.
• Knowledge base: It incorporates corporate documents, web pages, videos, audios, data files, and connected applications into the search context, ensuring that responses are based on internal content.
• Actions and MCP: They connect to pre-defined applications, custom interfaces, Webhooks, and MCP services, enabling the Agent to retrieve data or carry out business actions.
• White-label Portal: Combines multiple agents and pages to create a branded product center; it is possible to use the platform’s hosted address or a custom domain name.
• Access control: It supports public, member-based, and invitation-only access groups, and it is possible to set limits on the range of tools available, usage levels, upgrade options, and prices for different users.
• Built-in sales: Supports subscription, pay-per-use or points-based payment, one-time purchase, as well as external payments; payments within the platform are processed via Stripe.
• Users and Analysis: Manage member information, memory data, conversation records, and activity data, as well as monitor customer usage and Agent costs.

How Agent Builder works

Input, output, and multimodal capabilities

• Text and forms: End users can chat, as well as fill in structured fields such as short answers, long answers, multiple-choice questions, or checkboxes.
• File input: Builders can allow end-users to upload files and control the context budget allocated to those files.
• Voice input and playback: Voice input can be enabled, and the responses from the voice playback agent can be selected.
• Image generation: The image generation feature can be enabled by selecting an appropriate model; the actual cost varies depending on the model used and the amount of usage.
• Reasoning suggests that some agents can display the reasoning steps to users, but when it comes to sensitive workflows, it is necessary to first assess whether it is appropriate to make them public.
• Output format: The results can be used as chat replies, form responses, web page experiences, messages via various channels, API responses, or external actions.

Knowledge base support content

Scanning documents enables text recognition, and the content in Google Drive is synchronized in near real time; the content of the entire site is reprocessed only when changes occur.

Knowledge Explorer allows one to view the segmentation, relevance, and token consumption, making it useful for determining whether inaccurate answers stem from the search process or from the content itself.

Creation, deployment, and sales processes

1. Register an account and create a Workspace, determining whether the tool is intended for internal users, customer projects, or for public sale.
2. Start with a template or let the AI Helper create an Agent based on your requirements, then check the prompt, the model, and the capability switches.
3. Include authorized knowledge content, and use real-world questions to test retrieval and answering in Knowledge Explorer and Preview.
4. Configure Actions, Webhooks, or MCP tools to add confirmations and permission restrictions for high-impact actions such as writing, making payments, and sending messages.
5. Choose between Portal, direct link, web embedding, messaging channel, or API deployment, and adjust the brand, domain name, and access method.
6. Create public, member-only, or invitation-based access groups, and set parameters such as the tools visible to users, the number of times they can use those tools, points earned, and the paths for upgrading.
7. When a fee is required, connect to Stripe, select a subscription, one-time payment, or pay-as-you-go option, and view the final price page.
8. After going live, monitor the model’s credibility, user activity, processing fees, incorrect answers, and tool usage, then continuously adjust the model and its knowledge base.

Deployment method

Five web embedding styles

• Inline Flex: A responsive chat interface embedded within the page’s content area.
• Floating Action Button: It can be opened using the floating chat button located in a corner of the page.
• Popup from Button: A dialog box appears after clicking a custom button.
• Chat Input: Only one separate input field is provided, suitable for compact pages.
• Inline iframe: Use an isolated frame to place tools of fixed size.

Commercialization and end-user billing

• Subscription: It is possible to set subscriptions on a daily, weekly, monthly, or annual basis, and to configure different tools and quotas for various access groups.
• Per use or per credit: \"Uses\" counts one time for each user input, while \"Credits\" reflects the actual cost of the model; only one of these options can be chosen for a single Workspace.
• One-time payment: A fixed access fee is charged upfront, with the option to also purchase additional services on a pay-as-you-go basis to cover the ongoing costs of the model.
• External payment: It redirects the user to another payment page; the permissions after payment need to be managed by the developer themselves.
• Offers: It is possible to sell an individual agent, page, or folder separately, without having to create a complete package for each item.
• Platform processing fee: 10% for Gold plans, 8% for Pro plans, while the Business plan entails a retention rate of 98% on the income generated.

When end-users use the Agent, it consumes the model credits of the creator; therefore, the pricing should cover the costs associated with the model, platform processing fees, payment service charges, support costs, and taxes.

Prices and packages

The Gold and Pro annual subscription pages indicate a 20% discount; the price of these plans does not include any additional costs for model credits, payment processing fees, or external services.

The Business summary states that a customized credit limit is provided, while the comparison table on the same page mentions unlimited credit; these two statements are contradictory. Before signing the contract, it is necessary to have the sales representative include details regarding the credit limit, bandwidth, and any charges for exceeding those limits in the contract.

The official website provides a free entry point, but the current main price list does not clearly indicate the model credits, Workspace options, knowledge base resources, and deployment limitations associated with free accounts.

Model credit and additional costs

• One dollar worth of Pickaxe credits corresponds to the cost of using the model; each time an answer is generated, those credits are consumed.
• The credit for the Gold and Pro packages is refreshed every 30 days and does not accumulate; any unused monthly credit is not carried over to the next period.
• Extra credit purchased separately does not expire, while the credit included in the package is used before the additional balance.
• Actions and knowledge base searches do not directly deduct credit; the model usage is calculated only when the Agent generates the final output.
• Different models, context lengths, file and output lengths result in varying costs per interaction; it is necessary to first estimate these costs using actual traffic data.

API, CLI, MCP, and open-source status

Official organizations have public code repositories, but the core services at present are hosted products, and no open-source license applicable to the entire platform has been identified.

Copyright, data training, and commercial usage rules

• The builder retains the intellectual property rights to its own prompts, as well as ownership of the responses collected and returned through its Pickaxe.
• The results generated by the model are owned by the organization that created it, within the limits set by the terms of the model provider and as permitted by law.
• Publishing an Agent or Portal as a platform template grants other users irrevocable, royalty-free, and sublicensable rights to copy and use it; caution is required before doing so.
• Public agents and their locations of public embedding can be used for platform promotion; sensitive customer projects should employ appropriate privacy and access settings.
• The builder is responsible for informing end-users that the builder and Pickaxe support staff may view interactions when applicable.
• The interactive inputs and outputs from non-Pro accounts may be used to train the Pickaxe model, whereas documents uploaded to the knowledge base are explicitly not intended for training purposes.
• Data generated during the valid period of a Pro membership are not used for training, but data from before the upgrade or after it is terminated is not covered by this commitment.

Privacy and security

• The platform collects information such as accounts, payments, user inputs, files, communications, devices, logs, usage patterns, and domain names of embedded websites.
• The text submitted by end-users to a particular Pickaxe can be accessed and used by the owner of that tool; the creator must provide clear privacy information.
• The privacy policy does not specify a fixed duration for data retention; it only states that data is kept as necessary for providing services, ensuring security, handling disputes, making improvements, conducting training, and meeting legal requirements.
• The terms specify that static data is encrypted using AES-256, while transmitted data is encrypted with TLS; in addition, minimum privilege principles, vulnerability management, monitoring, and regular security tests are implemented.
• The price page indicates whether compliance with SOC 2, CCPA, and GDPR requirements is achieved; nevertheless, companies should still assess the control measures in light of their own type of data, location, and contractual obligations.
• The platform states that it does not sell personal information nor shares it for cross-context advertising; however, certain service providers and under specific legal circumstances, data may still be obtained.
• Users can apply, depending on their location, to access, correct, delete, restrict, object to, or transfer their personal information.

Refunds, cancellations, and risk warnings

• The terms of service state that purchases of Gold, Pro, additional plans, pay-as-you-go credits, and other paid features are final transactions and cannot be refunded.
• Monthly or annual subscriptions are automatically renewed until the user cancels them.
• Cancellation or downgrading takes effect at the end of the current billing period; the corresponding benefits can still be used during the already paid period.
• No proportional refund or credit is provided for partial payment terms, unused time, unused credit, or downgrades.
• Only verified major service disruptions or technical failures may warrant an exception decision by Pickaxe, which may also result in a deduction from the payment.
• The Beta feature is available as it is; no guarantees are provided regarding its official release, continuous availability, or data integrity. Critical operations should not rely on unreleased features.

Suitable for users and their capacity limits

• Individuals and organizations who are suitable for packaging courses, consulting methods, industry knowledge, or internal processes into paid AI products.
• It is suitable for AI advisors and agents who need to deliver white-label agents, customer portals, and web embeds quickly.
• It is suitable for teams that wish to gradually expand their capabilities from a no-code interface to development methods using APIs, CLI, and MCP.
• Model outputs can still be incorrect, and knowledge bases as well as strong prompts cannot replace fact-checking, human approval, and risk control.
• Complex Actions may read from or modify data in external systems; therefore, permissions should be restricted, parameters should be verified, and audit logs should be kept.
• In addition to platform fees, there are also model credits, processing fees, Stripe fees, as well as costs related to external applications and support; the business model must take all these factors into account.
• Models and channels may not be available in different regions; it is necessary to verify supplier restrictions and local compliance requirements before serving international customers.

Summary

The core value of Pickaxe lies in creating a complete, product-ready workflow that encompasses the creation of agents, the management of knowledge, deployment, access, and revenue generation – rather than merely producing a chatbot.

When making a choice, it is important to compare the cost of the actual model, the sales and handling fees, the rules for data training, the refund policies, as well as the target distribution channels, before deciding between the Gold, Pro, or enterprise version.