Xolving Labs

We join your team and get AI agents into daily use.

Xolving Labs is a forward deployed engineering company. We work inside your team to find the work AI agents can take on, build those agents into your repositories and tools with Claude Code and the Claude Agent SDK, and hand them over so your engineers run them. We also build the platform underneath, on Kubernetes, Terraform and CI/CD.

Get in touchSee services

Remote from Iași, Romania, with working hours that cover the full business day in Europe and on the US East Coast.

Agent working

Staging costs too much. Right-size the node group. Right-size the staging node group

The agent is working on the task. It opens a pull request when it is done. Opened by the agent for the task: Staging costs too much. Right-size the node group.

  1. read CloudWatch, 14 days of node metrics CPU p95 23%, memory p95 41%
  2. read envs/staging/eks.tf 6 nodes, m6i.2xlarge
  3. edit envs/staging/eks.tf - instance_types = ["m6i.2xlarge"]+ instance_types = ["m7i.xlarge"]- desired_size = 6+ desired_size = 4
  4. run terraform plan 0 to add, 1 to change, 0 to destroy
  5. run tflint, checkov passed
  6. open pull request #214 Right-size the staging node group

Approval gate. The agent cannot merge its own pull request. Waiting for a maintainer. The agent cannot merge its own pull request. Approved by you. CI takes it from here.

  1. merge #214, approved by you
  2. run CI applies the reviewed plan 1 changed
  3. done Staging runs on 4 nodes, m7i.xlarge about $1,300 a month less at on-demand prices

Ready

A scripted example of the agents we build with your team. Real runs use read-only cloud credentials, and only a person can merge.

Services

  • Forward deployed engineering

    We help your team adopt AI agents, from the first use case to daily use.

    • Finding the first tasks worth handing to an agent, together with your team, starting from the work it repeats every week.
    • Agent workflows on Claude Code and the Claude Agent SDK, with MCP servers for your own tools and data.
    • A harness around every agent: scoped permissions, hooks that block destructive commands, a token budget per run, and write access only through pull requests.
    • Deployment as short-lived Kubernetes Jobs with least-privilege credentials. Secrets stay out of prompts and logs.
    • Langfuse tracing, an audit log of every tool call, and alerts on cost and failure rates.
    • Documentation and a handover, so your team can run and change the agents without us.

  • Platform engineering

    The platform your developers ship on. We build it inside your team and can keep running it.

    • Kubernetes on EKS and AKS, or HashiCorp Nomad, Consul and Vault.
    • Terraform and Helm for infrastructure, ArgoCD for delivery.
    • CI/CD on GitLab CI, GitHub Actions or Azure DevOps, with the slow and flaky steps fixed.
    • Monitoring and alerting with Prometheus, Grafana and Alertmanager.
    • Smaller cloud bills across development, test and production.

    Watch an agent right-size staging

  • Site reliability engineering

    Migrations and day-to-day operations, with agents taking the routine work.

    • Migrations from on-premises to AWS, with reusable Terraform modules, a cost estimate per component, and Well-Architected documentation.
    • Cluster patching, security updates, and dependency updates through Renovate.
    • Incident response where an agent investigates the alert and drafts the root-cause report before the on-call engineer opens a laptop.
    • A monthly retainer, priced per month and not per token.

    Watch an agent triage an alertWatch an agent patch a CVE

Selected work

We don't publish client names.

  • Two AI products, from zero

    Cloud infrastructure built from scratch for an AI video-surveillance product used in office buildings, and for an aviation SaaS.

  • An on-premises to AWS migration framework

    Reusable Terraform modules, CI pipelines, custom tooling, per-component cost estimates and Well-Architected documentation, for a software company becoming an AWS partner.

  • Multi-agent DevOps tooling

    Agents on the Claude Agent SDK that run as ephemeral Kubernetes Jobs. They open pull requests and draft manifests, and have no write access to production.

  • Lower infrastructure cost

    Cloud spend cut across development, test, acceptance and production, and across the CI/CD runner fleet.

How we work

  • Embedded

    Part time or full time in your standups, tools and on-call rotation. Billed monthly.

  • Fixed-scope project

    A written scope and a fixed price before work starts. Suits a first agent pilot, a migration or a platform build.

  • Monthly retainer

    Patching, updates and on-call after a project ends. Billed monthly.

What happens after you get in touch

  1. A 30-minute call

    You describe the problem. We say whether we're the right fit.

  2. A written proposal

    Scope, timeline and price, before any work starts.

  3. Delivery

    Pull requests in your repositories, reviewed by your team.

  4. Handover

    Documentation and a walkthrough, so your team runs it without us.

Tell us what you're working on

A few lines about the problem is enough. We reply from info@xolvinglabs.com.

We use what you send only to reply to you. See the privacy policy.