Accompanying code for my Nextflow Summit 2026 talk on using Microsoft Foundry to orchestrate Nextflow workloads.
As bioinformatics workflows grow in scale and complexity, automating the data retrieval, analysis, and interpretation is imperative for scalable analyses. What if pipelines could reason about their own execution, adapt to intermediate results, and collaborate with other agents to drive scientific discovery? In this session, we introduce best practices in building multi-agentic workflows for bioinformatics that combine the easy AI agent capabilities provided by the Microsoft Foundry platform with the power of Nextflow. We will demonstrate how to design AI agents that can interpret biological intent, dynamically parameterize and launch Nextflow pipelines, monitor execution state, and post-process results, closing the loop between human hypothesis and computational analysis.
This talk is a bio-themed deep dive into content from “Building Agentic Solutions with Microsoft Foundry” by Colby T. Ford, Ph.D. (O’Reilly Media, 2026) and is intended for bioinformaticians and AI developers who are exploring how agentic systems can augment reproducible computational biology.
Navigate to your Seqera Cloud account and generate an API token.
In Microsoft Foundry, create a SeqeraCloud connection using the Custom Model Context Protocol (MCP) tool and configure it with your Nextflow token.
- Remote MCP server endpoint:
https://mcp.seqera.io/mcp - Authentication: Key-based
- Credential name:
Authorization - Credential value:
Bearer <your Seqera cloud token>
Create a bioinformatics-eval-agent agent in Microsoft Foundry using the agents/bioinformatics-eval-agent-prompt.md prompt. (Look at the agents/bioinformatics-eval-agent.yaml as a reference.)
This agent evaluates a given user input and breaks it down into structured components that can be used by subsequent agents. This include an evaluation of the biological problem, data requirements, workflow steps, etc.
Create a seqera-agent agent in Microsoft Foundry using the agents/seqera-agent-prompt.md prompt. (Look at the agents/seqera-agent.yaml file as a reference.)
This agent communicates with the Seqera Cloud MCP server to search for datasets, discover nf-core pipelines, and executes workflows based on the structured plan provided by the bioinformatics-eval-agent.
Create a new Workflow by using the provided workflow/nf-Workflow-001.yaml YAML file. You'll need to connect the agent nodes to the specific agents that you create (if their names differ from the names in the YAML file).
Once you create the Workflow, click the ▶ Preview button to see the workflow graph and a chat panel. Here's an example prompt:
Help me to design a bioinformatics workflow for SRA bioproject PRJNA784582.
