About cookies on this site Our websites require some cookies to function properly (required). In addition, other cookies may be used with your consent to analyze site usage, improve the user experience and for advertising. For more information, please review your options. By visiting our website, you agree to our processing of information as described in IBM’sprivacy statement. To provide a smooth navigation, your cookie preferences will be shared across the IBM web domains listed here.
Blog post
Get to know watsonx Orchestrate
What developers need to know about watsonx Orchestrate
Watsonx Orchestrate is an agnostic multi-agent orchestration platform designed to help organizations scale the deployment of AI agents across business domains. It can connect with various enterprise applications, automation tools, and third-party AI models, allowing it to integrate seamlessly with existing business systems, providing a single, user-friendly interface for building and interacting with AI agents.
At the core of watsonx Orchestrate's developer experience is the Agent Development Kit (ADK), a pro-code, Python-based toolkit and CLI designed for building, testing, and deploying AI agents. The ADK gives you full control over the agent lifecycle, enabling you to define tools, configure behaviors, manage connections, including Model Context Protocol (MCP) for standardized connections, and register knowledge sources directly in code.
With the ADK, you can easily prototype agents in a containerized environment, debug their interactions, and seamlessly deploy them into production. Agents built with the ADK can be deployed across multiple environments, including IBM Cloud, AWS, and your local machine, all managed through a single, cohesive workflow.
While the ADK gives developers deep programmatic access, the platform also exposes this power through a visual, intuitive layer: the Agent Builder. The Agent Builder is a no-code environment built on top of the ADK, designed to make agent creation accessible to non-developers and business users alike, without sacrificing technical flexibility. It allows users to:
- Define agent profiles and domain behaviors
- Connect to external tools, APIs, or internal services
- Upload and configure domain knowledge, or link to external sources
- Deploy across multiple channels (Teams, WhatsApp, web, etc.)
- Test and debug agents in a live, interactive preview
Agents created in the Builder are powered by watsonx.ai foundation models and follow an agentic reasoning loop:
- Perception - understanding input and context
- Reasoning - planning via prompt strategies
- Action - invoking tools, triggering workflows, or collaborating with other agents
You can also enrich agents with:
- Knowledge grounding - via document upload or connection to vector databases (e.g., Milvus, elasticsearch)
- Custom tools - defined through OpenAPI or Python
- Collaborator agents - imported from your org or imported from other environments
- Behavioral instructions - to govern how agents respond and act
Here’s a quick tour around the Agent Builder and its main features:
Together, the Agent Builder and ADK offer a unified development stack that meets users where they are, empowering business users to design functional agents with no code, while giving developers the flexibility to build, extend, and manage agents programmatically.
Development can also be accelerated by leveraging pre-built AI agents from the watsonx Orchestrate catalog as templates, making it easier to customize and adapt functionality for specific business needs. This approach helps teams reduce time-to-value and quickly scale AI adoption across the organization.
By combining an approachable visual interface with a robust pro-code toolkit, watsonx Orchestrate enables businesses to democratize agent development while still enforcing standards, scaling securely, and integrating seamlessly into existing infrastructure.
What’s new in the June 30 release
The latest release brings several key enhancements aimed at giving developers greater control, broader reach, and improved orchestration fidelity. Watch the release video for a quick peek at some of these new features:
- Agent reasoning modes: Default vs. ReAct
- New channel integrations: WhatsApp & Teams
- Connections page: External app management
- Voice configuration (Beta)
- ADK native evaluation
Agent reasoning modes: Default vs. ReAct
You can now explicitly select how your agents think:
- Default: A streamlined tool-usage mode where the LLM executes instructions directly via tool invocations. Use Default for more deterministic, prompt-based tasks.
- ReAct: A reasoning-forward strategy that uses intermediate steps, chain-of-thought generation, and planning before action. Use ReAct when transparency, step-by-step rationale, or multi-hop logic is needed.
New channel integrations: WhatsApp & Teams
Agents can now interact with end users via:
- WhatsApp (via Twilio integration)
- Microsoft Teams (direct connection)
This opens up new channels for integrating AI agents directly into enterprise communication stacks.
Connections page: External app management
A new Connections UI allows users to configure integrations with third-party platforms, enabling agents to securely interface with enterprise APIs and backend systems without hardcoding. This is the new entry point for managing:
- OAuth / API keys
- External tool mappings
- Authentication layers
Voice configuration (Beta)
Voice-based interaction is now in beta for testing inside the Agent Builder:
- Speech-to-text and text-to-speech via IBM watsonx APIs
- Available in the chat preview window for prototyping multimodal agents
This feature is ideal for devs exploring hands-free or accessibility-first use cases.
ADK native evaluation
A new Agent Evaluation Framework is coming to the ADK to help developers test, measure, and improve agent behavior through simulated interactions called trajectories.
The Agent Evaluation Framework includes tooling to help users generate and scale ground-truth datasets that reflect real use cases. By comparing agent interactions against this reference data, including expected tool calls, conversation context, and final summaries, developers can determine whether agents are completing tasks accurately and effectively.
This framework enables a structured, iterative development loop that brings greater testability, observability, and control to agent development in watsonx Orchestrate
Ready to build?
watsonx Orchestrate is more than an automation tool, it’s a platform for building and orchestrating agentic systems from multiple frameworks that reason, collaborate, and scale across the enterprise. With powerful pro-code and no-code development options, growing extensibility, and now built-in evaluation capabilities, it gives teams everything they need to move fast without losing control.
As the ecosystem continues to expand, with new tools, templates, and ways to integrate across platforms, developers and business users alike can shape how intelligent agents transform their workflows. Whether you're building with the ADK or configuring agents in the UI, watsonx Orchestrate is ready for what’s next.
Ready to start building? Review the watsonx Orchestrate docs, and get started with your free trial today!