Magpai

What is Magpai?

Magpai was an AI workspace operated by the Australian company Magpai Pty Ltd; it was developed under the leadership of its founder, Ben Skinner. The product evolved from a multimedia node-based creation tool to a platform that incorporates knowledge bases, AI assistants, business integrations, and automated workflows.

Its goal is to enable non-technical teams to gather information scattered across emails, documents, project management tools, and customer systems into a single workspace, after which the Maggie assistant can retrieve relevant context, generate content, and carry out tasks across different applications.

Current operating status

• Magpai ceased operations in 2026; its current homepage is a farewell page for the founder’s products, and it no longer provides options for registration, logging in, or making purchases.
• The farewell page indicates that the product’s lifecycle spans from 2023 to 2026, and it clearly states that Magpai will no longer be in operation.
• Old pricing pages, documentation, update logs, and legal pages may still be found through searches, but these are materials from before the service was discontinued and do not reflect the current availability of the service.
• The availability of the servers related to old workflows, integrations, and APIs cannot be confirmed; therefore, no new production processes should be created based on them, nor should any further payments be made.
• At this stage, it is more appropriate to consider Magpai as a product that is no longer in use; original users should prioritize exporting their data, revoking any third-party authorizations, and switching to alternative tools.

Core historical functions

Maggie’s team AI assistant

• Research and retrieval: It allows searching on the public web as well as accessing information stored within a team’s knowledge base; it is suitable for researching competitors, locating relevant materials, and answering internal questions.
• Writing and summarization: It takes in questions, documents, or business context to generate emails, articles, summaries, and structured explanations, with the ability to adjust the wording according to team terminology.
• Tool invocation: After connecting to a business application, Maggie can read or update external data within the conversation, rather than merely providing text responses.
• Continuous tasks: Longer tasks can be processed in the background; previous versions support reminder functions, as well as management of conversations based on their status – in progress, pending approval, completed, or error.
• Team context: Brain, dictionaries, and custom skills are used to store knowledge, terms, and rules for repetitive tasks, enabling different team members to work in a more consistent manner.

Brain Knowledge Space

• Users can organize notes, links, files, and other business-related content in visual canvases and folders to provide the AI with context about the team.
• Links allow for the creation of titles, descriptions, and image previews, which helps in identifying the content within the canvas; notes support collaborative updates and automatic layout.
• Maggie’s responses can be saved as notes, and the results of the workflow can also be stored in Brain in the form of components, thus creating a closed loop that covers retrieval, execution, and archiving.
• A knowledge space does not automatically ensure that the data is accurate or up-to-date; teams still need to manage permissions, update outdated information, and check whether the AI is using incorrect context.

Code-free workflows

• The visual editor allows users to drag and drop steps and view the connections between them, making it suitable for combining form inputs, prompts, integration actions, approval processes, and notifications into a workflow.
• Workflows can be run manually, or they can also be executed automatically according to schedules described in natural language; they are suitable for weekly reports, data synchronization, and periodic reminders.
• Each workflow is assigned a dedicated email address; the content of the emails can be parsed as form inputs or prompt context, and after execution, a notification indicating completion or an error is sent.
• The results of the execution can be shared through public pages; users can regenerate the sharing credentials or disable sharing. When sensitive information is involved, the scope of public access should be restricted.
• In the past, workflow notifications supported email, SMS, and Slack, and could alert the relevant persons when approval was required, tasks were pending, or tasks had been completed.

Business integration and collaboration

• The historical documents list over forty different integrations, covering categories such as Microsoft, Google, Slack, project management, CRM, notes, and time tracking.
• Public update records have confirmed support for connections to services such as OneNote, Outlook, Monday.com, ClickUp, Harvest, and Meta ads; the specific read and write permissions vary depending on the service.
• Attio once provided an official application developed by Magpai, which enabled two-way synchronization and allowed CRM data to be utilized in business processes.
• Multi-user conversations, team switching, and participant management enable members to collaborate with Maggie within the same context.
• For third-party integration, it is necessary to grant permissions to external accounts; after the service is discontinued, one should go to the security settings of each connected service to revoke the Magpai permissions.

Historical workflow

1. Create a team space and invite members who need to share knowledge, workflows, and integrations.
2. Organize documents, notes, links, and team dictionaries in Brain to determine which materials can be made available for use by AI.
3. Connect the required business applications from the team settings, and check individually the read, write, and management permissions for Magpai requests.
4. Describe the task to Maggie, or combine input, prompts, integration actions, approval, and notification steps in the visual editor.
5. Run a trial with non-sensitive samples to check field mapping, generated results, external object writing, failure handling, and cost consumption.
6. Set up manual approval processes, scheduling, usage limits, and notification methods, and then make the streamlined process available for use by the team.
7. Regularly check the operation logs and access permissions, and disable any public sharing, workflows, or third-party authorizations that are no longer needed.

The above are the typical methods used before shutting down a service. It is not possible to carry out this process using the product at present; it can only serve as a reference for redesigning automation processes during migration.

Input, output, and applicable tasks

Historical applicable users and scenarios

• Small and medium-sized enterprise teams: They consolidate the context of multiple business applications, thereby reducing the need for team members to search for information across different systems repeatedly.
• Operations and project teams: generate weekly reports on a regular basis, keep tasks up to date, send reminders, and require manual approval before making any key changes.
• Sales and Customer Team: Reads CRM data, emails, and meeting details to prepare follow-up actions or update customer records.
• Content and design team: Early versions allowed for the processing of multimedia elements such as images, videos, documents, and presentations by combining them using nodes.
• Developers and automation consultants: Workflow integration into internal systems is possible through APIs, Webhooks, or Python bindings, but the existing methods for connection are no longer suitable for use.

Historical prices and the impact of service disruptions

• The Pro package card on the old pricing page stated that a monthly allowance of $20 was included, but the FAQs on the same page said it was $50; these two statements are contradictory.
• Any usage of AI that exceeds the allotted quota was charged separately based on actual consumption, and limits could be set per user, team, or workflow.
• The old page allows for upgrades or downgrades at any time; once canceled, the associated benefits remain valid until the end of the current billing cycle.
• The old terms stipulated that no refund would be given for the remaining period of a prepaid subscription once it was terminated, but consumers’ rights and the rules regarding specific payment methods might still take precedence.
• As the service has been discontinued, the amounts mentioned above are only for historical reference and should not be used as current quotes or purchase recommendations.

API, SDK, and open-source status

• The Business package has historically provided Webhook and API access, allowing workflows to be triggered from third-party applications with results returned thereafter.
• PyPI once released a Python binding for Magpai; the last version known to have been published was in 2023, and it required the use of a Magpai API key.
• This Python package is licensed under the MIT license, but this applies only to the bundled packages; it does not mean that the Magpai web platform, servers, models, or the entire product are open source.
• There are no publicly available GitHub repositories that have been confirmed to be still under maintenance, and no official code or licenses that would allow one to deploy a complete Magpai service on their own after it stops operating.
• It is not possible to determine whether the original API services, keys, Webhook addresses, and examples will continue to function; therefore, production systems should eliminate any reliance on them and replace the relevant credentials.

Considerations regarding data, privacy, and copyright

• The old provisions stated that the data entered by users, along with any intellectual property associated with it, remained the property of the users, who were responsible for ensuring the legitimacy, accuracy, and quality of such data.
• The user has granted the operator permission to use, copy, transmit, store, and back up data for the purpose of providing services and related activities.
• Third-party integration providers have access to data within the scope required for interoperability, and the operator is not responsible for any data disclosure, modification, or deletion caused by these providers.
• The privacy policy states that personal information may be stored by Google Firebase in locations outside Australia, and it may also be processed by contractors and service providers.
• Historical security measures have included two-factor authentication and restricted roles, but the policy also states that internet transmissions cannot be guaranteed to be completely secure.
• Google user data can be deleted upon request; for databases that are in use, deletion takes place within 30 days after identity verification is completed. Accounts that have not been used for 12 months are notified first before being deleted or anonymized.
• The old regulations required that requests for data copies be submitted within 30 days of termination, and stated that the data in the service would be permanently deleted after 60 days of termination; exporting data might incur fees under certain conditions.
• After shutdown, the actual capabilities regarding response handling, backup deletion, and export cannot be verified; the original user should keep records of the requests made and revoke any authorizations granted to external applications.

Capacity boundaries and migration recommendations

• Research, summaries, and business decisions generated by AI assistants may be incorrect, and they cannot replace financial, legal, HR, or security approvals.
• Cross-application automation amplifies the impact of erroneous fields, incorrect objects, and excessive permissions; important write operations should involve manual verification as well as rollback mechanisms.
• Sharing workflow results publicly may expose customer, employee, or internal business information; it is necessary to review and revoke old sharing permissions before making any migration.
• Workflows rely on external interfaces and model services; changes in third-party permissions, fields, or prices can all lead to the failure of the process.
• The original user should export the data that is still accessible, document the workflow logic, retest it on the alternative platform, and rotate the API keys, Webhook keys, and connection tokens.

Summary

Magpai combined team knowledge, the Maggie assistant, code-free workflows, and business integrations to create a unified AI workspace; its feature is that it enables non-technical users to build automation through conversations and visual steps.

The product is no longer in use, and the old documents and pricing information can only help in understanding its past capabilities. For existing users, it is not important to try to register again; rather, what matters is to complete the migration and cleanup of data, permissions, credentials, and related processes.