Practical overview
When it fits
Use CubePlex when a team needs an agent whose workspace, knowledge and working state persist across tasks, especially when tool connections and access need organization-level governance.
Who it is for
- Teams seeking a shared agent that can continue work across tasks.
- Organizations that need to curate MCP connectors and govern agent access to tools.
- Teams evaluating self-hosted agent infrastructure on Docker Compose or Kubernetes.
Documented uses
- Maintain a stable team agent role, shared knowledge and working state across tasks.
- Enable approved MCP tools for a workspace using organization-curated connectors.
- Package reusable workflows as skills and make them available in the appropriate workspace.
- Run scheduled or webhook-triggered agent tasks and deliver versioned artifacts in conversations.
- Continue workspace work in isolated sandboxes whose files and working trees persist across restarts.
Requirements & limits
The repository is public on GitHub.
- Workspace data and agent configuration are team-owned and persist across tasks.
- Sandbox storage persists across restarts, including files, packages and the working tree.
- MCP access is organized through an administrator-curated connector catalog and workspace enablement.
- Credentials can be scoped to organization, workspace or user, using OAuth or static credentials.
- Docker Compose and Kubernetes are the stated self-hosting options; the README lists Python 3.12+ and Node 20+.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Create a CubePlex account and start an AI conversation
Set up an account, ensure a model is available, and send a first message in CubePlex.
Before you start
- A CubePlex instance installed and a URL provided by its administrator.
- An email address and password; if email verification is enabled, access to the inbox for the one-time code.
- At least one AI model enabled. Organization owners or admins need a provider API key to add a provider and enable models.
Steps
Open the CubePlex URL and register with your email and password. If prompted, enter the verification code. Complete the onboarding wizard to name the organization, choose a slug, and create a workspace; the first user becomes the organization owner.
If no model is available, an organization owner or admin opens Admin > Models > Model Providers, adds a provider such as Anthropic or OpenAI with its API key, and enables models for the team. Members can use models their admin has enabled.
Choose New conversation, select a model from the selector above the chat input, enter a message, and press Send.
Expected outcome
CubePlex returns a response in the conversation. Depending on the model and available tools, the response may also include tool calls, code execution results, or artifacts.
Guide limitations
- CubePlex Cloud is described as coming soon; the documented account flow directs users to a self-hosted installation until then.
- The first user becomes the organization owner; subsequent users join the existing organization as members.
- Members cannot add model providers through the documented member flow; they use models enabled by an admin.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 15 October 2026
- Original website title
- cubeplexai/cubeplex · Official homepage title · retained in its source language