The tools we reach for. The ones your team keeps.

What every project rests on.

• Next.jsType-safe full-stack framework. We ship most web work here.
• TypeScriptNon-negotiable. If it doesn't type-check, it doesn't ship.
• Postgres + pgvectorRelational data and vector search on the same system.
• Drizzle / PrismaType-safe ORMs so the schema stays the source of truth.
• React + TailwindUI that stays consistent as teams and features grow.

The layer that makes agents safe to run.

• OpenAI / AnthropicModel choice is workflow-driven, not brand-driven.
• LangfuseAgent traces, evals, and cost visibility in one place.
• MCPTool contracts that scope what an agent can actually do.
• Temporal / InngestDurable workflows for long-running, human-in-the-loop agents.
• Eval harnessesCustom-built per workflow. No model change ships without one.

What you keep when we leave.

• OpenTelemetryAgent, service, and human decision traces — on one spine.
• GrafanaDashboards your team can actually read at 2am.
• SentryError surface with agent-aware context tagging.
• Audit logging patternsEvery decision, tool call, and handoff — explainable on demand.

How the work gets to production.

• Vercel / Firebase App HostingWhere the front doors live.
• GitHub ActionsCI with gates: typecheck, build, eval, deploy.
• TurborepoMonorepos that stay fast as they grow.
• Feature flags + rollbackEvery agent-touching release has a kill switch by default.