Forge an intelligent, autonomous AI workforce
Manhattan Agent Foundry™ simplifies the creation of custom supply chain commerce AI agents that work exactly how, where and when you need them.
See how easy it is to AI
Manhattan Agent Foundry makes it easy to build custom agents that harness the full potential of Manhattan Active solutions. Quickly build agents with specialized skills and tools, transforming and automating work throughout your supply chain commerce processes.
Multiple ways to build AI agents
Manhattan Agent Foundry empowers both business practitioners and developers to introduce new agents that are skilled, trusted, and high-performing, digital members of your workforce.
.webp/Zz1iOTRmNThhOGY4YTYxMWYwYTYyZTA2YTRmOWFkNjMzZQ==&width=528&height=297)
From AI
Develop an agent by having a natural language conversation with Agent Foundry so it can create an agent for you that meets your needs.
From Template
Simply start with an existing agent and, in plain English, describe the changes you want. Helper agents and tools take care of the rest, building the new, ready-to-deploy agent.
.webp/Zz1lMTJmMDk3OGJjMDUxMWYwYWVmY2VlYjZjOGZjOGJkZA==&width=528&height=297)
From Scratch
Manage the complete process of building new agents, including fine-grained control over reasoning, business logic (i.e., workflows), and tool interactions (i.e., APIs).
Easier and faster AI agent creation
Manhattan Agent Foundry is a powerful, intuitive, no-code application for building supply chain commerce agents with speed and efficiency.
Simplify
Visual
Provides a highly intuitive visual and declarative agent development environment that eliminates the complexities of traditional coding, enabling users to rapidly and safely launch new agents with unparalleled ease.
Streamline
Unified
Manhattan Agent Foundry is a unified, consistent environment combining deterministic business logic and tools (i.e., API calls) with generative AI reasoning (i.e., prompt definitions and LLM calls), streamlining agent development.
Accelerate
Composable
Offers a composable design that accelerates the development of new agents by leveraging a comprehensive library of reusable components, including existing agents, skills, and tools.
Secure
Controlled
Manhattan Agent Foundry automatically enforces guardrails that control agent behavior, allowing users to create safely and at scale. For example, API guardrails ensure that builders can only access pre-defined services for authorized systems.
Modify and extend AI Agent tools with Manhattan ProActive™
Use a visual development environment to update data, APIs, and UIs in Manhattan Active solutions to support your AI agents.
Advanced AI agent development
Manhattan Agent Foundry provides the essential building blocks to create and manage highly capable agents that work across supply chain commerce processes.
Reasoning
Incorporate LLMs at any stage in a workflow using prompt templates with instructions provided in natural language.
Workflows
Implement intricate, multi-step workflows by defining tasks and using conditional logic to guide the flow of execution.
Tools
Select, define, and invoke Manhattan Active Platform APIs, external APIs, foundational agents, and external agents, allowing agents to interact with backend systems of record.
Data & Content
Create, modify, and remove variables and parameters within the agent's memory to manage state and context.
Communications
Call internal or external agents with standard APIs. Expose agents as callable tools, to be discovered and orchestrated by Google Gemini Enterprise.
Collaboration
Call internal or external agents with standard APIs. Expose agents as callable tools, to be discovered and orchestrated by Google Gemini Enterprise.
Guardrails
Implement procedural guardrails using conditional workflow logic to enforce business rules and require human approvals where necessary.
Deployment
Publish to an agent marketplace for discovery and use or register internal “foundational agents” as callable tools for other agents to use.
Lifecycle
Manage the entire agent lifecycle, from creation and versioning to retirement, within a single, unified framework.
Share your AI Agents with Google
AI agents created Manhattan Agent Foundry can be published to Agent Marketplace within Gemini Enterprise to accelerate integration and collaboration with agents from other solutions across the enterprise.
Collaborate with other AI agents
Manhattan Agent Foundry supports MCP (model context protocol) and A2A (agent-to-agent protocol) standards to coordinate actions, exchange data, synchronize workflows, and streamline coordination.
MCP
Governs the transport and orchestration layer of how messages move between agents and systems, focusing on message control and reliability.
A2A
Defines how two or more autonomous agents communicate, cooperate, and coordinate tasks directly with one another in a distributed environment, focusing what is being communicated.
Frequently Asked Questions
Artificial intelligence is rapidly advancing, and one of its most promising frontiers is agentic AI.
Manhattan Agent Foundry is included with all Manhattan Active solutions to build, modify, and manage your AI agent workforce.
At the forefront of innovation and usage of agentic AI in supply chain commerce, our mission is to help customers improve decision accuracy and enhance performance possibilities.
With Manhattan Agent Foundry’s dual-path agent development approach, both business practitioners and professional developers can create new agents.
For business practitioners without coding knowledge, the platform offers a build-by-template method. Users can express their instructions in plain English, such as asking an agent to check load utilization during planning. The technology will then automatically suggest relevant pre-built “helper agents” from its component library to facilitate the creation of the new agent.
Advanced users and developers have the option to take it further by building agents from scratch. Utilizing the advanced development features of Manhattan Agent Foundry, developers can undertake more complex tasks. This includes fine-tuning reasoning prompts for specialized scenarios, selecting Manhattan Active APIs or other data sources, and defining multi-step workflows.
This dual-path approach provides both business-friendly simplicity and developer-level control, allowing our customers to innovate at their own pace.
Unlike many solution providers, Manhattan does not restrict AI agent-building capabilities to simple applications, such as chatbots or query-based agents. With Manhattan Agent Foundry, customers can create workflow-based agents equipped to handle complex tasks. The agentic AI framework within Manhattan Active Platform leverages microservices and an orchestration engine to ensure that new agents operate safely within our solutions. This robust development environment makes Manhattan Agent Foundry far more flexible and comprehensive compared to other offerings in the market.
Manhattan Agent Foundry puts trust and security at the forefront. Multi-layered safeguards and controls are automatically enforced throughout the AI agent development process. These measures are configured by administrators and builders, ensuring the enterprise-level control and confidence needed to scale AI agents effectively. For instance, instructional guardrails guarantee that each agent operates under a strict set of system-level instructions, ensuring it is grounded in real data and preventing biased or fabricated responses. Additionally, content sanitization controls offer customizable rules to clean both the data sent to and received from LLMs. This helps prevent prompt injection and ensures clean and properly formatted responses to users.
Ready-to-deploy Manhattan Active Agents, and custom agents, utilize numerous tools to perform their work. These can be Manhattan Active solution APIs that are called upon to access and manipulate data or control processes within a solution, as well as APIs from any external system. Additionally, external agents can be integrated using MCP and A2A protocols. Agents can also be published to Google Gemini Enterprise, streamlining connection to AI agents from external systems. Manhattan Agent Foundry’s trust and security features ensure agents invoke only predefined and approved service definitions. These definitions encapsulate the connection, authentication, and request structure for backend systems, guaranteeing they only access data and perform actions within explicitly authorized systems.
Manhattan Agent Foundry and Manhattan ProActive are complementary tools included with all Manhattan Active solutions. Agent Foundry focuses on the development of AI tools, while Manhattan ProActive is designed to extend core solution capabilities (e.g., data, UI, APIs) to meet new requirements. For instance, if a new data attribute is needed to support an automated task carried out by an AI agent, Manhattan ProActive can be used to add the attribute. Then, the attribute becomes accessible to the agent through the appropriate API of the Manhattan Active Platform.
Realize the benefits of a unified platform
Learn more about how Manhattan Active Platform and AI Agents can unify your supply chain commerce operations.