watsonx Orchestrate ADK
The watsonx Orchestrate ADK is the Python package and command line tool that builds and ships IBM watsonx Orchestrate agents from a terminal instead of a web form, and it is where an IBM i shop attaches a Model Context Protocol (MCP) toolkit so an agent can reach Db2 for i. It installs with pip install ibm-watsonx-orchestrate, carries an MIT licence, and its toolkit command accepts either a local MCP server or a remote web address plus a transport, which is exactly what IBM's open source IBM i MCP Server serves.
What is the watsonx Orchestrate ADK?
Think of watsonx Orchestrate as a kitchen. The web interface is the front of house, where you fill in forms to build an agent. The ADK is the service door at the back, where everything an agent is made of arrives as files and typed commands. ADK stands for Agent Development Kit. IBM's own documentation calls it a set of tools designed to make it easy to build and deploy agents using IBM watsonx Orchestrate, packaged as a Python library and command line tool. The GitHub repository puts it shorter still: the command line client for watsonx Orchestrate's agent builder experience.
Two things arrive in the same install. One is a command named orchestrate, which imports and manages the pieces an agent is made of: the agent itself, its tools, its toolkits, its knowledge bases and its connections. The other is watsonx Orchestrate Developer Edition, which IBM describes as a local server on your computer, giving you a dedicated development environment. For an IBM i shop that second piece is the useful one. You can build an agent and point it at your own data on a machine you control, before anything goes into a shared environment.
What does it need to run?
Memory IBM asks for on the computer running Developer Edition
Figures from IBM's Developer Edition setup page, which lists an 8 core processor in every case and does not state a disk size. GB means gigabyte. The same page says the Developer Edition no longer depends on an external container engine, so an older note about it running under a Docker container is out of date.
The kit does not include a model. IBM's setup page says you need access to at least one inference provider: watsonx Orchestrate on software as a service, watsonx.ai, Groq, or a provider of your own through the AI Gateway. The entitlement itself comes from a watsonx Orchestrate account, including the free 30 day trial, or a myIBM key for an on premises licence. You start the local server with orchestrate server start and an environment file, and orchestrate server purge removes it along with its data.
How do you attach an MCP toolkit?
Four steps, following IBM's own tutorial
- Install the kitpip install ibm-watsonx-orchestrate on Python 3.11 or later. IBM's tutorial also had Node.js on the machine, tested at version 23.7.0, because many MCP servers are Node programs.
- Start the local serverorchestrate server start reads an environment file and brings up watsonx Orchestrate Developer Edition on your own computer.
- Add the toolkitorchestrate toolkits add registers the tools and gives them a name and a description. The description is not optional, because the agent uses it to decide when to reach for the tools.
- Open the chatorchestrate chat start opens the agent, and the imported tools appear in its tool list, ready to be chosen.
Steps follow the IBM Developer tutorial published December 17, 2025 by Ahmed Azraq and Moises Dominguez Garcia, checked against the ADK toolkit documentation. MCP stands for Model Context Protocol, the open standard for handing an AI agent a set of tools it can call.
What the toolkit command asks for, local server against remote server
| Question | Local MCP server | Remote MCP server |
|---|---|---|
| Where is it | A folder on the same computer as the kit, holding the server package | A web address you supply |
| How does it start | A command you give, such as a Python or a Node command | Already running, so there is nothing to start |
| What else is required | requirements.txt for a Python server, package.json for a Node server | A transport, either SSE or Streamable HTTP |
| Which tools arrive | A named list, or a star for all of them | A named list, or a star for all of them |
SSE stands for Server-Sent Events. IBM's remote toolkit page adds two limits worth knowing: watsonx Orchestrate waits up to 30 seconds for the server to answer with its tool list, and only key value connections can pass credentials. A toolkit can also be the python kind rather than the mcp kind. Table rows are not linked because this site's content blocks render as plain text.
Can it reach IBM i?
Yes, but you supply the bridge. IBM ships no IBM i or Db2 for i connector for watsonx Orchestrate, so nothing in the kit knows what a library list is. What the kit does have is the remote toolkit form, and it asks for only two things: a web address and a transport. IBM's own open source IBM i MCP Server answers exactly that shape. Its quick start runs the server on port 3010, answering at /mcp over HTTP transport, and it installs from npm as @ibm/ibmi-mcp-server. Give the toolkit command that address and the SQL tools written in the server's YAML files become tools an agent can choose. SQL stands for Structured Query Language, and YAML is a plain text layout for settings.
Two cautions. A local import points at a folder on the computer running the kit, so the server process lives there, not on the IBM i; for a server already running on or beside your IBM i, the remote form is the one that fits. And the address in the quick start is a development default on the local machine. A real setup needs an address Orchestrate can reach and the sign-on choices the IBM i MCP Server documents, which include OAuth. IBM also publishes a documentation MCP server for the kit itself, at developer.watson-orchestrate.ibm.com/mcp, and says IBM Bob detects and connects to it automatically.
How long has it been shipping?
The kit's public record, start to current release
- March 28, 2025IBM creates the public repository, IBM/ibm-watsonx-orchestrate-adk, written in Python under an MIT licence.
- April 30, 2025Version 1.0.0 lands on PyPI, the Python package index, as ibm-watsonx-orchestrate.
- November 25, 2025Version 2.0.0 opens the second major line.
- November 11, 2026Version 2.20.0 ships, with 119 releases on record and 179 stars on the repository.
Dates read from the GitHub repository record and the PyPI release data for ibm-watsonx-orchestrate on November 16, 2026. A release count shows pace, not stability; the move from the 1 line to the 2 line took under seven months.
Related Terms
watsonx Orchestrate
IBM's AI agent and automation platform for orchestrating multi-step business workflows across enterprise applications. Since August 2026 it can launch IBM Bob from its Create an agent dialog through a Build with Bob preview, but it still ships no IBM i or Db2 for i connector.
watsonx Orchestrate Agent Builder
watsonx Orchestrate Agent Builder is where an AI agent is actually defined: its name, its written instructions, the large language model that reasons for it, and the list of tools it is allowed to call. For an IBM i shop the tool list is the whole story, because it accepts a Model Context Protocol (MCP) toolkit, and that is the practical way to point an Orchestrate agent at Db2 for i data when IBM ships no IBM i connector.
Model Context Protocol
An open standard, released by Anthropic on November 25, 2024, that defines a client-server architecture for connecting AI applications to external data sources, tools, and prompts over JSON-RPC 2.0. IBM's IBM i MCP Server is an implementation of it for Db2 for i and other IBM i resources.
IBM i MCP Server
An open-source, IBM-published server implementing the Model Context Protocol (MCP), letting AI agents and LLMs query Db2 for i data and take defined actions on IBM i using natural language. In beta since October 2025.
ContextForge MCP Gateway
IBM's open source gateway that puts one governed endpoint in front of every Model Context Protocol server, so an IBM i shop can register its Db2 for i tools once and let any AI client reach them with shared login checks, rate limits and logging. IBM's own IBM i MCP Server project ships a container stack that runs the gateway on port 4444 beside an IBM i SQL tools server on port 3010.
Db2 for i
The relational database management system built into every IBM i operating system, tightly integrated with IBM i rather than a separately installed database product.