stackai

What is StackAI?

StackAI is a platform for building AI agents and workflows tailored for enterprises. Teams can connect large models, internal knowledge, business systems, and manual approvals within a visual interface, thereby transforming individual queries into repeatable business processes.

The product highlights its capabilities in enterprise-level governance, evaluation, and deployment; it is suitable for creating internal knowledge assistants, as well as for document analysis, compliance checks, customer service, and executing tasks across different systems.

Product status after joining Asana

Asana completed the acquisition of StackAI in May 2026. According to the official statements from both parties, StackAI will continue to operate as an independent product and brand; existing customers do not need to migrate their services, nor are they required to integrate them with Asana.

• The existing workflows and platform will continue to function as usual, and users do not need to take any additional actions as a result of the acquisition.
• The existing data connectors, model connections, and third-party integrations will remain in place.
• Local deployment and virtual private cloud deployment remain part of the enterprise solutions.
• StackAI’s solution engineers and AI strategy team continue to provide support for enterprise implementations.
• Users can still create agents that are not related to Asana; Asana is just one of the business systems that can be connected to them.

Main functions

• Visual workflow: Connects inputs, models, knowledge bases, tools, logic, and outputs through nodes.
• Corporate knowledge base: Import files, cloud storage contents, and web materials to provide a controlled search context for answering queries.
• Multi-model support: Different model providers can be selected based on the task, with the administrator able to restrict the available models.
• Agent tools: Enable AI to query business systems, create tasks, call interfaces, or execute multi-step processes.
• Human intervention: Review, approval, or additional information is added at key stages to reduce the risks associated with automatic execution.
• Evaluation and analysis: Observe the operation results, errors, and usage patterns to provide a basis for subsequent iterations.
• Multi-channel distribution: It can be deployed as a chat interface, form, website assistant, message bot, or backend service.
• Corporate governance: Offers permissions, access control, logging, model management, and options for deployment environments.

Visual workflow editor

StackAI’s Flow canvas breaks down complex applications into individual nodes. Users can configure prompts, models, search functions, conditional logic, code, and external tools, and then use connections to determine the flow of data.

Which processes are suitable for implementation?

• Read a set of contracts, extract the clauses, and generate review summaries based on risk rules.
• Receive customer inquiries, retrieve answers from the knowledge base, and pass complex issues on to human agents.
• Analyze sales data, update customer records, and create follow-up tasks.
• Compile multiple reports, compare key metrics, and generate management briefs in a fixed format.
• Combine forms, files, and business system data into a pre-approval check process.

Knowledge base and retrieval-enhanced generation

The platform can organize corporate documents and data from connectors into a knowledge base, allowing agents to first retrieve relevant snippets before generating answers. Compared to including the entire document in the prompt, this approach is more suitable for larger sets of data.

• It supports uploading common office documents such as PDF, Word, and PowerPoint.
• Data can be loaded from Google Drive, Notion, and other enterprise data sources.
• Citations can be enabled for answers to help users return to the relevant source for verification.
• Data from different departments can be separated into individual projects or knowledge bases, thereby reducing the risk of mixed permissions.
• Companies still need to define rules for updating, deleting, versioning, and access permissions, in order to prevent old data from continuing to affect the responses.

Models and multimodal capabilities

StackAI is not tied to any single large-scale model; the official documentation lists models or hosting services such as OpenAI, Anthropic, Google, Meta, Mistral, xAI, Azure, and AWS Bedrock. Companies can also connect to local model endpoints based on their own requirements.

Model governance

• Administrators can control the model providers and specific models that are allowed to be used within the organization.
• The team can choose a model based on accuracy, speed, context, cost, and data policies.
• Different workflows can use different models; it is not necessary to assign all tasks to the same supplier.
• For sensitive items, it is necessary to determine where the data will be sent, whether it will be retained, and what service terms apply.
• After switching models, the evaluation needs to be run again; it cannot be assumed that the output behavior will remain exactly the same.

Connectors and business tools

The platform offers hundreds of enterprise integration options for reading knowledge, querying structured data, or performing various operations. The number of available nodes varies depending on the version and enterprise configuration, so it is necessary to check this information before starting a project.

• Collaboration and project management: Asana, Slack, and other team work tools.
• Customers and sales: Customer relationship management platforms such as Salesforce.
• Content and documents: Sources such as Google Drive, Notion, and corporate document repositories.
• Data and analysis: data warehouses, databases, and structured query nodes.
• Communication channels: website chat, Slack, WhatsApp, text messages, and other means of interaction.
• Custom systems: Connect internal enterprise applications through interfaces or code nodes.

Manual approval and secure execution

Automated agents may carry out errors, which is why StackAI allows for human intervention in the process. Teams can have the AI prepare suggestions or drafts first, which are then approved by a supervisor before being implemented.

• Pause the process before performing high-risk actions and request manual approval.
• Forward results with low confidence or insufficient evidence to business experts.
• Manually fill in the missing fields before allowing the workflow to continue.
• Record the approver, the outcome of the processing, and the reasons for any abnormalities to facilitate subsequent audits.
• For external transmissions, as well as for making changes to funds, permissions, and sensitive records, full automation should not be the default setting.

Evaluation, monitoring, and version management

When moving from a demonstration environment to a production environment, companies need to assess whether the agents can accomplish their tasks reliably, rather than merely focusing on a few well-presented responses. StackAI offers capabilities for performance analysis and management of the development lifecycle, which are useful for testing, comparing, and iterating workflows.

• Create a test set that covers normal, boundary, rejection, and attack scenarios.
• It measures search hits, factual accuracy, format compliance, and task completion rate separately.
• Compare the results before and after changes to prompts, knowledge bases, models, and tool configurations.
• Monitor latency, failures, call volumes, and the proportion of cases requiring manual intervention.
• Before releasing a version, it should be tested with a small group of users first, and then its use should be expanded gradually.

Publication method

Deployment and data environment

Safety and compliance

The StackAI website lists the company’s capabilities related to SOC 2 Type II, GDPR, and HIPAA, and offers options such as SSO, access control, dedicated environments, and business partner agreements. Compliance certifications cannot replace the customer’s own risk assessments.

• Limit users’ access to knowledge bases, tools, and models based on their roles and projects.
• Use enterprise single sign-on and offboarding processes to manage the lifecycle of accounts.
• Before formal integration, check the data processing protocols, sub-processors, and retention policies.
• Classify and minimize medical, financial, legal, and personnel data.
• Establish mechanisms for log review, anomaly alerts, backup, recovery, and response to security incidents.
• If deployment is carried out using a local or virtual private cloud, it is also necessary to clarify the boundaries of responsibilities for operations and maintenance between the two parties.

Applicable industries and scenarios

• Financial services: Processing research materials, investment memorandums, due diligence, and compliance pre-checks.
• Healthcare and life sciences: Organizing documents, policies, and operational knowledge under access control.
• Industrial enterprises: Analyze technical documents, maintenance records, procurement records, and operational reports.
• Business operations: Link cross-departmental documents, forms, and system tasks into traceable processes.
• Customer service: Answers questions based on a knowledge base, and routes complex or sensitive requests to human agents.
• Sales and Marketing: Researching customers, organizing meeting details, updating records, and creating follow-up tasks.
• Legal and Risk Management: Extract contract clauses, compare policies, and prepare a draft for risk assessment.

Which users are it suitable for

• Teams that wish to quickly test enterprise agents but do not want to write all the orchestration code from scratch.
• Large organizations that need business experts and engineers to work together to design processes.
• Enterprise clients that place emphasis on model selection, permissions, auditing, evaluation, and manual approval.
• Operations teams that need to connect to various knowledge bases, collaboration systems, and business data sources.
• Organizations that need to choose between cloud, dedicated environments, virtual private clouds, or on-premises deployment.

StackAI Usage Guide

1. Choose a task with clear boundaries and verifiable results, such as extracting fixed fields from a contract.
2. Create a project and configure the input nodes as well as the models to be used on the canvas.
3. Upload test materials or connect to a knowledge base, and set the search range and citation methods.
4. Add prompt words, conditions, code, or external tool nodes to connect the complete data flow.
5. Prepare real anonymized samples to check facts, format, refusal responses, and abnormal branches.
6. Add manual approval, permissions, and error handling, then choose to publish via chat, form, or API.
7. After going live, continuously monitor operation records and user feedback to carry out version updates.

Enterprise launch process

1. Goals, responsibilities, and prohibited actions are determined jointly by personnel in business, technology, security, and compliance.
2. Track the sources of data, model providers, external tools, and the destination of the data at each step.
3. A gold-standard test set is established, with thresholds set for accuracy, completion rate, latency, and cost.
4. First, run it in read-only or draft mode to test it; once stability is confirmed, enable write operations.
5. Manual approval is required for high-impact actions, and procedures for suspension, rollback, and manual replacement are in place.
6. Choose between the cloud, a dedicated environment, a virtual private cloud, or on-premises deployment, and complete the security review.
7. After a limited rollout, observe any abnormalities and the adoption rate, then gradually expand the scope of users and data.

Prices and packages

As of this verification, StackAI’s official pricing page offers two options: Free and Enterprise. The free version costs $0 per month, while the Enterprise version comes with a customized quote; the specific benefits and quotas are specified on the purchase or contract page.

Key benefits of the enterprise version

• All functions and data loaders are available; the specific scope of access is determined by the contract.
• Options for dedicated infrastructure, virtual private clouds, or on-premises deployment.
• Access control, single sign-on, and enterprise security capabilities.
• Dedicated solution engineers and higher-level support services.
• The number of runs, seats, and other capacities are configured according to the enterprise’s requirements.

What needs to be taken into account when billing?

• The definition of Run affects the 500-free-use quota and the estimation of enterprise capacity.
• It is necessary to confirm with sales regarding the costs associated with model invocation, voice services, external services, and cloud resources.
• Dedicated environments, on-premises deployment, implementation, and solution engineering services may be charged separately.
• The size of the knowledge base, concurrency levels, number of seats, data connections, and level of support all affect the pricing offered by a company.
• Prices and benefits may change; when making purchases, it is necessary to refer to the current pricing page and the contract.

Product advantages

• Bring models, knowledge bases, tools, and logic together on a visual canvas to facilitate cross-functional collaboration.
• It supports multiple model providers as well as local models, thereby reducing the reliance on a single model.
• It features a wide range of enterprise connectors that enable the conversion of responses into business actions.
• It offers manual approval, operational analysis, and governance capabilities, bringing it closer to a production-grade agent platform.
• Covers the cloud, dedicated infrastructure, virtual private clouds, and on-premises deployment.
• The free version can be used for low-cost testing, while complex corporate projects benefit from the support of an engineering team.

Usage restrictions and precautions

• Visualization does not mean zero learning cost; complex processes still require an understanding of data, models, and exception handling.
• There is no fixed public price for the enterprise version; it is necessary to conduct a thorough analysis of the costs associated with operation and implementation before making a purchase.
• The connectors and models come from different suppliers, and their permissions, stability, and data policies are not entirely consistent.
• Even with enhanced retrieval, incorrect citations, missed citations, or outdated information may still occur; key conclusions must be verified manually.
• Intelligences capable of performing write operations will amplify errors; therefore, the least possible permissions along with human approval are required.
• Local deployment can improve data control, but it increases the responsibilities related to upgrades, monitoring, capacity, and maintenance.
• The long-term direction for product integration following the acquisition may continue to change, and companies should pay attention to official updates.

The difference between StackAI and ordinary chatbots

GitHub and the open-source status

StackAI has an official GitHub organization and makes some of its repositories and examples available publicly, but the complete front-end, back-end, governance, and hosting services of its enterprise platform are not released as a fully open-source product. Just because there is public code available does not mean that StackAI can be considered an open-source platform.

Basic information

Recommendation score

4.5 / 5. StackAI achieves a good balance between visualizing intelligent agents, enterprise connectors, multi-model governance, human-machine collaboration, and flexible deployment, making it suitable for organizations that place emphasis on practical implementation; however, the pricing offered by the company is not transparent, and complex processes still require joint management by engineering, security, and business teams.

Frequently Asked Questions

What is StackAI mainly used for?

It is used to integrate large models, corporate knowledge, business systems, and manual approval processes into AI agents and workflows that can operate continuously.

Can StackAI still be used independently after being acquired by Asana?

Yes. According to the official statement, StackAI will continue to operate as a separate product and brand; existing customers do not need to migrate or integrate it with Asana.

Is StackAI free?

There is a free plan available; the official options include 500 executions per month, 2 projects, and 1 seat. For enterprise-level services, custom quotes are provided.

Is programming required for StackAI?

The basic workflow can be constructed using visual nodes, but complex logic, system integration, security, and production operations still require technical expertise.

Can I connect my own data?

Yes, it is possible to upload common office documents or connect to cloud storage, knowledge management tools, and other corporate data sources, which can then be utilized by intelligent search functions.

Which large models are supported?

It supports multiple major model providers as well as locally configured model endpoints by enterprises; the list of specific models will be updated alongside platform updates.

Can it be deployed privately?

The enterprise version offers options for virtual private clouds and on-premises deployment; the specific architecture, scope of responsibilities, and pricing need to be confirmed with the authorities.

Is StackAI suitable for medical data?

The official website lists the enterprise capabilities related to HIPAA as well as the options for business partner agreements, but customers still need to carry out their own compliance reviews and make the necessary configurations.

Does StackAI provide APIs?

It is possible to integrate the workflow as a backend capability into enterprise systems or custom interfaces; however, authentication, rate limiting, and error handling must be handled separately.

Is StackAI an open-source platform?

It is not a fully open-source platform. While there are some public repositories on the official GitHub, the core commercial components, as well as the enterprise governance and hosting functionalities, remain closed source.