Agent platforms
Agent runtimes, MCP servers, and structured tool interfaces. Agents that edit real configuration and code, not a generic filesystem chat.
Software and systems
AI agents, on infrastructure that can carry them.
HoopSoft designs agent platforms and the Kubernetes, Terraform, and security layers underneath. From MCP tooling and agent orchestration to production clusters that stay up when those agents are left running.
Capabilities
Agents in delivery
Orchestration, LLM-assisted review, and security audits inside a normal engineering workflow. Agents as part of the pipeline, with the same gates as everything else.
Cloud native platforms
Kubernetes, Terraform, Flux, and self-hosted PaaS as code. Right-sizing, GitOps, and providers teams can apply without a snowflake control plane.
Security engineering
Hardening across the stack: path traversal, injection, secret handling, and supply-chain issues in the tools agents and clusters already depend on.
Where we spend our time
Runtimes and MCP
Contributions to agent assistants and protocol tooling: session behavior, startup UX, and MCP servers for precise file edits (JSON, YAML, TOML, markdown, AST).
Workload right-sizing
In-place pod resource adjustment. Vertical scaling without the usual restart tax, built for clusters that already run production traffic.
PaaS as code
Providers and modules for self-hosted application platforms. Resources, data sources, and acceptance tests so apply is repeatable.
Contract-first systems
Specification-driven tooling where the intended behavior is written down first and checked, including work on an AI-native contract language.
Upstream hardening
Fixes landed in widely used projects: browser automation, TLS stacks, Terraform, GitOps controllers, and C library edge cases on WASM.
Languages and platforms
Java and Spring, Go, Rust, TypeScript. Kubernetes, AWS, Azure, Flux, and CI that can run an agent without becoming the incident.
About HoopSoft
HoopSoft is a software company focused on agentic systems and the infrastructure they require. The practice covers enterprise Java through Kubernetes-native platforms, and now centers on AI agents that are safe to operate in production.
Engagements typically mix platform engineering, security review, and agent integration. The same bar applies to all three: it has to run, someone else has to be able to review it, and if a check cannot be completed the work stops instead of continuing unsafe.
Practice areas
- Generative AI and LLM platforms
- Agent frameworks and MCP
- MLOps and AIOps
- Kubernetes application platforms
- AWS architecture and security
- Application security
- Terraform and GitOps
- Java / Spring systems