Here is what the validation covered, why we think the data layer is the part of agentic AI that decides whether agents reach production, and what changes for AWS customers.
In November 2025, AWS expanded its AI Competency with three agentic categories: Agentic AI Applications, Agentic AI Tools, and Agentic AI Consulting Services. AWS called it its largest Specialization launch to date, with 60 launch partners. In announcing the categories, AWS cited IDC data that 65% of organizations expect full deployment of agentic AI by 2027.
The Agentic AI Tools category is for the infrastructure layer: the platforms teams use to build and operate agents, rather than the agents themselves.
An agent that can reason but cannot reliably read the CRM, the warehouse, the ticketing system, or the contract repository does not get far. In most enterprises, that data sits across hundreds of SaaS applications, databases, files, and APIs. Each source has its own schema, its own access rules, and its own compliance requirements.
That is the problem Nexla is built for. The platform gives agents and AI applications a governed path to enterprise data:
AWS states that partners in the agentic categories go through technical validation and must show successful customer implementations. For Nexla, that meant:
AWS approved the application on October 27, 2026.
Instacart, the largest online grocery marketplace in North America, has used Nexla since 2017. Each retail partner sends product and pricing data in its own format, and custom integration for each one had become a bottleneck as the partner base grew.
With Nexla’s automated data engineering, Instacart supported a 748% increase in retail partners, cut retailer launch time by up to two months, and saw about 5X less data engineering maintenance work. Read the case study.
Nexla makes sharing data between companies, in any format, really easy,” said Brandon Leonardo, Co-Founder of Instacart.
The same pattern applies to agents: the work that slows them down is usually connecting, normalizing, and governing the data underneath, not the model.
Most agent projects stall at the same point: the agent works in a demo, then hits the reality of hundreds of enterprise systems with different schemas, access rules, and compliance requirements,” said Saket Saurabh, CEO and co-founder of Nexla. “Our focus has been giving agents governed, reliable access to that data. Achieving the AWS AI Software Competency in Agentic AI Tools means AWS has reviewed our architecture, security, and AI governance, and AWS customers now have a validated way to connect their agents to enterprise data.”
The AWS AI Software Competency recognizes partners validated by AWS. In November 2025, AWS added three agentic categories: Agentic AI Applications, Agentic AI Tools, and Agentic AI Consulting Services. The Agentic AI Tools category is for the infrastructure layer: platforms, frameworks, and development tools that help organizations build, deploy, and manage agentic AI systems. Partners go through technical validation and must show successful customer implementations.
AWS completed a technical review of Nexla’s architecture, security controls, and AI governance practices through its validation checklist. Nexla also submitted four enterprise customer case studies demonstrating production deployments. AWS approved the application on October 27, 2026, and Nexla is now listed as an AWS AI Software Competency Partner in the AWS Partner Solutions Finder.
Yes. Nexla’s AWS Marketplace storefront is live. For eligible customers, purchases through AWS Marketplace may count toward existing AWS spending commitments. Teams can also find Nexla as a validated AWS AI Software Competency Partner in the AWS Partner Solutions Finder.
Context is concentrated know-how, often more sensitive than the data it describes. Why it needs task-specific delivery, two identities per call, and lineage.
One LLM chat can’t know every system. Task-specific MCP servers scope data to each job, so answers are accurate and traceable.
How a context engine grounds, assembles and delivers context by task: 7 responsibilities, a task model, typed tools over MCP, and lineage you can inspect.
