Case study · 01
Pluto
AI agents that work alongside your team: a secure control plane for provisioning and operating tenant-specific agent deployments.
Days
8
1
What we built
Pluto is a commercial product: a secure AI agent that connects to the tools a business already uses and automates the workflows their team does manually. No custom software build-out: an agent deploys in days, works alongside the team, and stays inside the business's own boundaries.
Under the hood, Pluto is a control plane for provisioning and operating tenant-specific agent deployments, where every customer gets their own isolated agent runtime rather than a shared bot with shared context.
How it's built
A monorepo of eight applications around one core idea: isolation per tenant.
- Control plane
- A SvelteKit + Convex dashboard that owns product policy, control-plane state, provider credentials, and deployment orchestration.
- Agent runtime
- Built on the Eve agent framework: durable chat sessions, tools, and skills served over HTTP, deployed per tenant.
- Inference ledger
- A standalone service (Effect + Postgres) that meters and bills every model call, so usage is accountable to the penny.
- Channels
- An iMessage relay, for example, so agents meet teams in the tools they already live in.
- Companion surfaces
- Desktop shell, mobile app, admin tooling, and a public docs site.
Why it matters
Pluto is the productized version of what Goshen Labs does in services: workflow-first agents, strict tenant isolation, metered and auditable usage, and deployment in days rather than quarters. If you want the pattern applied to your organization's workflows, that's an engagement we know how to run, because we've already run it for ourselves.