Platform Overview

The Octo platform is the cloud-native, multi-user counterpart to the standalone editor: the same visual editor, backed by a control plane that stores integrations in Postgres and deploys each one as its own Kubernetes workload. Where the standalone editor reads flow files from a local directory, the platform gives a team shared storage, versioning, one-click deployments, and central operations.

Signing in lands you on the dashboard — an overview of your integrations and their live deployments:

What the platform adds

• Multi-user sign-in. OIDC single sign-on (Auth.js) gates the whole app. Each teammate gets a durable user identity, integration edits are attributed to the user who made them, and write operations can be restricted to specific roles.
• A persistent integration store. Integrations live in Postgres — uniquely named and organized into folders — instead of files on one person's disk.
• Version tags (snapshots). Tagging an integration freezes its definition and a copy of its resources under an immutable name. Deployments always ship a tag, never the live working copy.
• One-click deployments to Kubernetes. The orchestrator turns a tag into a Kubernetes Deployment of the generic octo-runtime image, with a stable internal URL per deployment and an optional public HTTPS endpoint on its own subdomain.
• Resources and secrets. Env files and templates are authored per integration and frozen with each tag; sensitive values live in a cluster-wide secret store and are bound to deployments by reference, never stored in the database.
• A clustered object store. A deployment's replicas share a versioned, deployment-scoped key/value store served by the orchestrator, with encrypted-at-rest secret namespaces — plus an object browser API for inspecting stored values.
• Central logs. Every runtime pod ships structured logs over NATS to a log aggregator that persists them to Postgres, queryable across all deployments.
• API keys and MCP. Per-user bearer keys (and OAuth access tokens) authenticate programmatic clients, including the platform's MCP endpoint for AI-assisted authoring.
• Live status. Deployment status streams from Kubernetes informers through NATS to the browser as Server-Sent Events, so you watch pods come up in real time.

The pieces

Two services make up the control plane: the platform app (apps/platform, a Next.js app that serves the editor and fronts everything as a backend-for-frontend) and the orchestrator (orchestrator/, a Go HTTP API that owns the database and is the only component that talks to the Kubernetes API). Architecture walks through how they fit together.

Running the platform

The platform ships as a Helm chart with images for every component. See Deploying with Helm for the chart, GCP with Terraform for a complete single-VM reference deployment, and Local k3d for a development cluster.