A cloud platform for developers and agents.
Multi-tenant isolation, a unified control plane, and a first-class agent protocol. Five products on one platform, billed by the minute, managed end to end.
Convoy Cloud started with a question: what does a cloud platform look like when it's designed today, not inherited from 2014? We built the answer from scratch on modern open-source infrastructure with the operational defaults that should have always been standard: real tenant isolation, automatic scaling, managed TLS, built-in observability, and a full audit trail from day one.
Then we built the part nobody else has. ACP, the Agent Control Protocol, exposes the entire platform as a programmable surface. Agents can deploy services, spawn sandboxes, tail logs, inspect health, scale resources, and tear them down without a human in the loop. Not a dashboard with an API bolted on. A native interface for autonomous software.
Developers get a clean dashboard, a CLI, and Git-push deploys. Agents get ACP. Same control plane, same isolation, same billing. Two interfaces, one cloud.
Three architectural layers define Convoy Cloud. Everything above them (apps, functions, databases, sandbox, observability) is a product that runs on this foundation.
Multi-tenant virtualization
Every workspace runs inside its own isolated tenant boundary. Dedicated resource limits, network boundaries, scoped credentials, and a separate observability stack. Your workloads are invisible to other tenants, and theirs to you.
The control plane
One control plane orchestrates everything: capacity checks, builds, deploys, scaling, health monitoring, rollback, and audit logging. Whether a request comes from the dashboard, the CLI, or an agent, it flows through the same pipeline with the same guarantees.
Agent-native protocol
ACP (Agent Control Protocol) exposes the entire platform as a programmable surface. Agents can deploy services, spawn sandboxes, tail logs, inspect health, scale resources, and shut down workloads. Not a wrapper around a dashboard. A first-class interface to the cloud.
We run workloads on hardware we manage directly. No hyperscaler margin stacking, no opaque pricing layers. We control the metal, the networking, and the pricing.
Billed by the minute in LKR with a 10-minute minimum. Scale-to-zero by default. Agents can loop through dozens of experiments without turning every iteration into permanent spend.
Every workspace gets its own Grafana Org with fine-grained metrics, logs, and built-in alerting. Seven alert conditions are seeded on day one. Observability is a built-in primitive, not an add-on.
CVE tracking across all components, image scanning before anything runs, software bill of materials per release, and critical patches shipped in hours. We build on auditable open-source infrastructure because security comes from visibility.
Every product is reachable over ACP. Agents can deploy, observe, scale, and tear down resources without a human in the loop.
Per-minute LKR billing, scale-to-zero defaults, no idle charges. Cost is a design constraint, not an afterthought.
Workloads run on infrastructure we manage directly. Less margin stacking, more control over performance and pricing.
Every workspace gets its own tenant boundary, network policy, resource limits, and observability scope. Your data stays yours.
Push code, get a URL. Builds, deploys, TLS, DNS, and scaling are handled. Focus on the application, not the pipeline.
Full audit timeline, deployment history, real-time logs, and a public changelog. You see everything that happens on your account.
See the changelog. Read the roadmap. Or start building.
We ship in public. Everything we build lands in the changelog before it lands in marketing.