ParkMobile had installed Kubernetes. That should have been the triumphant part. The parking company was moving beyond its old infrastructure, putting a modern platform underneath a service used by millions. Yet its developers still needed a practical way into the machinery. The cluster existed. The bridge to everyday work was missing.
- Solo.io controls how applications, APIs and AI agents connect.
- Its customers buy enterprise features and support around open source software.
- The lesson: make complicated infrastructure easier for the next person to use.
Solo.io supplied an API gateway. ParkMobile’s platform team added abstractions, including Helm templates for routing, so developers could use the platform without becoming orchestration specialists. That division of labor explains much of Solo’s appeal. The company sells to people who must make complicated systems usable by other people. A successful installation is only the beginning.
“We didn’t want the developers to have to learn the ins and outs of everything related to Kubernetes.”
Anthony Mallica
Platform Engineering Tech Lead, ParkMobile
Sell the next useful piece
When Idit Levine founded Solo.io in 2017, she saw enterprises breaking large applications into smaller services. That created an awkward consequence: more software components meant more conversations between components. Someone had to decide where requests went, who could make them, and what happened when a destination stopped answering.
Levine’s early calculation was about timing. Service mesh adoption would take years. API gateways already had buyers. Building Gloo around Envoy gave Solo an immediate product and a route toward the larger networking problem. The name was a small joke with a practical point: Gloo would glue environments together.
Today, Solo Enterprise for kgateway manages API traffic using Kubernetes Gateway API. Solo Enterprise for Istio handles service-to-service connectivity, security and visibility. The older Gloo Gateway and Gloo Mesh names remain in its documentation. These are tools for platform engineers, infrastructure teams and security operators whose problems extend across applications, clusters and clouds.
For an application team, the practical controls are quite specific. Send a small share of traffic to a new release before committing everyone. Limit requests when a backend is overloaded. Give a team authority over its own routes while retaining shared security policy. These are the everyday decisions hidden inside the phrase application networking.
Solo’s distinction is its proximity to the underlying projects. CTO Louis Ryan co-founded Istio and gRPC. The company helped create Istio’s ambient mode, which removes the requirement for a proxy beside every application workload. That expertise matters when a customer’s question reaches beyond a configuration screen and into the behavior of the networking system itself.
The commercial arrangement is familiar but consequential: open source software attracts adoption; enterprise subscriptions add capabilities, hardened builds and support. Professional services cover work such as implementation and training. Against gateway alternatives such as Kong or Apigee, Solo’s pitch centers on Kubernetes-native connectivity and its connection to Istio. Customers can also operate community software themselves.
Open source entry. Tailored enterprise pricing.
Solo scopes estimates to environment and use case. Budget for integration, rollout and operation as well as the subscription.The slow route to a safer system
Confluent offers a useful corrective to the idea that buying infrastructure finishes the job. Its platform team needed consistent encryption and visibility as services spread across environments. Solo’s case study describes a rollout to more than 100 services over roughly a year, beginning with permissive settings before tightening to strict mutual TLS, which authenticates both ends of a connection.
Integration with Confluent’s SPIRE-based identity system proved difficult. Solo’s engineers helped troubleshoot and contribute upstream fixes. The reported outcome was mTLS across all mesh-enabled clusters. The copyable part is the sequence: establish visibility, resolve identity mismatches, then enforce stronger rules. Skipping straight to enforcement would miss the very dependencies the rollout needed to uncover.
Domino’s UK moved from deployment cycles of two to three weeks to days during its work with Solo.
Domino’s UK supplies another kind of evidence. Its 18-month modernization combined gateway and mesh software with a move from a monolith to microservices. Teams gained more independence, while timeouts, retries and circuit breakers helped manage failures between services. The useful outcome was a faster release rhythm. It belonged to an engineering program, rather than a software license alone.
The new caller is an agent
AI introduces a different participant into these conversations. An agent can choose a tool, call a model, and act on a user’s behalf. Now the network must carry more than a request. It must preserve the relationship between the user, the agent and the authority being exercised.
Solo’s Rust-based agentgateway handles model access, Model Context Protocol tools and agent-to-agent communication. Its controls include centralized provider credentials, token limits and tracing. Kagent supplies a Kubernetes-native runtime for agents. Agentregistry catalogs the agents, tools, skills and other artifacts teams might otherwise rebuild or lose track of.
The buyer’s question changes accordingly. Counting API calls tells you whether traffic is flowing. Following an agent’s tool calls tells you what it attempted, which identity it used, and where the model budget went. Operators need that detail to investigate an action after the fact.
The direction became clearer in September 2026. Solo added standalone deployment to its enterprise agentgateway, acknowledging teams running outside Kubernetes. Agentdesktop brought discovery, configuration and credential controls to employee machines. Then October’s agentshield announcement addressed infrastructure exposed to agents outside an organization’s control. Each extends the same question: who can reach what?
A smaller first step
There is money behind the ambition. Solo announced a $135 million Series C in October 2021, led by Altimeter Capital, at a $1 billion valuation. Those figures describe a financing event. The customer stories offer a more useful buying test: identify the connection that is difficult to operate, then decide whether Solo makes it easier.
For a small application with few services, adding a mesh may create more operational work than it removes. For a larger platform, begin with a bounded route or workload, assign an owner, and test the identity model before broad enforcement. ParkMobile’s templates and Confluent’s gradual rollout suggest a sensible standard: the next person should inherit less complexity than you did.