cocalc.ai

AI agents in CoCalc run inside the same persistent Linux project that holds your files, software, and running services, rather than in a separate sandbox. That means an agent can open, edit, and save the actual project files your collaborators already use, and you can review or take over its work afterward. CoCalc supports an integrated Codex agent, Claude Code (an experimental preview on sites that enable it), and other command-line agents run from project terminals.

The core idea: agents work where the project lives

A CoCalc project is a durable shared Linux environment. It keeps notebooks, terminals, code, documents, services, discussion, history, and recovery in one place. Agents are not given a copy of your work in an isolated container — they operate on the project itself.

This has three practical consequences:

• Direct file access. The agent reads and writes the same files you and your collaborators see. There is no export/import step between "agent workspace" and "real project."
• Access to tools and services. The software installed in the project and any running services are available to the agent, because it runs in that environment.
• Shared state. Because the project is persistent, the agent's changes remain in place and are subject to the project's history and recovery features.

How the agent options differ

The choice is mostly about which agent you prefer, not about where the work happens — all three operate against the same project.

What collaboration looks like

Collaborators see edits live. When an agent saves changes, others in the project can open what it saved, or take over the work themselves. This is the main difference from a standalone agent sandbox: there is no handoff step, because the agent was already working in the shared environment.

A typical flow looks like this:

1. An agent (Codex, Claude Code, or a terminal-based agent) makes changes in the project.
2. Collaborators watch the edits appear live.
3. Anyone can open the saved result and continue the work manually.

How this compares with an agent sandbox

An isolated agent sandbox gives the agent its own environment, separate from your real project. That can be useful for containment, but it means the agent's output has to be moved into your project before it's usable, and the agent can't touch your installed tools or running services.

In CoCalc, the agent works in the project's actual environment. The trade-off is the reverse: you get direct access to real files and services, and you rely on the project's history and recovery features to manage changes rather than on the sandbox boundary.

Where to go next

• To try it, start on CoCalc.ai and open a project.
• For agent-specific workflows, see the agent workflows page.
• For the sandbox comparison, see the compare-with-agent-sandboxes page.
• For deployment choices — hosted CoCalc.ai, local or single-VM, or a customer-operated private deployment — review the product paths.
• For cost and licensing details, see Pricing and licensing; for help, see support and sales.

CoCalc is developed by SageMath, Inc., and has offered collaborative computing online since 2013.

A CoCalc project is a persistent shared Linux environment that can hold notebooks, terminals, code, documents, running services, discussion, history, and recovery in one place. That means the same project can support coding, data analysis, writing, and collaboration without splitting work across separate tools. It is a good fit when you need long-running work that multiple people (or AI agents) can return to, and less useful if you only need a one-off script in a disposable sandbox.

What one project can contain

According to CoCalc's own description, a single durable project keeps these elements together:

• Notebooks — interactive computational documents.
• Terminals — command-line access inside the project.
• Code — files and software the project runs.
• Documents — written material alongside the code.
• Services — running processes your collaborators already use.
• Discussion — conversation attached to the project.
• History and recovery — a record of changes plus the ability to restore earlier versions.

Because these live in one shared Linux project, collaborators see edits live and can open what an agent saves, or take over the work themselves.

Workflows this supports

Software build and use with AI

CoCalc's headline use case is building and using software with AI. You can run the integrated Codex agent or Claude Code, or run other command-line agents in project terminals. The agents work where the project lives, so they use the same files, tools, and running services your collaborators use. Note that Claude Code is described as an experimental preview on sites that enable it, and it works with your personal Claude Pro or Max subscription.

Multi-person collaboration

Since files and the running environment are shared, a team can work in the same project rather than syncing copies. Live edits and shared services reduce the "which version is current" problem.

Long-term maintenance

History and recovery matter when a project runs for months. You can trace what changed and restore an earlier version instead of rebuilding from scratch.

How CoCalc runs

You are not limited to one deployment style. CoCalc lists three paths:

Pricing and licensing details are on CoCalc's pricing page; support and sales are separate contact paths. The site does not state that any tier is free, so check the pricing page before assuming cost.

When this fits — and when it doesn't

Choose a single CoCalc project when you need persistent, shared, recoverable work with agents and collaborators in the same environment. Consider a lighter sandbox instead if your task is short-lived, single-user, and you don't need history or shared services. If you need a private deployment, the customer-operated path is the relevant option to evaluate.

CoCalc offers three deployment paths: hosted CoCalc.ai, a local or single-VM installation, and a customer-operated private deployment. The right choice depends on who should control the infrastructure and how much maintenance you want to take on. Hosted is the fastest way to start; local/VM and private deployments shift control to you or your organization.

The three deployment paths

CoCalc is developed by SageMath, Inc., and has been offered as collaborative computing online since 2013.

Hosted CoCalc.ai

This is the default starting point. You use CoCalc as a hosted service, so there is no server to provision or maintain. Projects are persistent shared Linux projects that keep files, notebooks, terminals, history, and recovery together.

Choose this when you want the shortest path to a working project and don't need to control the underlying infrastructure. Note that some capabilities are gated by site configuration — for example, Claude Code is described as an experimental preview on sites that enable it, and it works with your personal Claude Pro or Max subscription.

Local or single VM

CoCalc can also run locally or on one VM. This keeps the environment under your control while staying relatively simple — a single machine or instance rather than a distributed setup.

Choose this when you want the project environment on hardware you manage, or when a self-contained instance fits your workflow better than a hosted service. You take on the setup and upkeep that hosting would otherwise cover.

Customer-operated private deployment

For organizations that need a private deployment they operate themselves, CoCalc offers a path to evaluate a customer-operated private deployment. Here the infrastructure and operation sit with your organization.

Choose this when control over where data and services live matters more than minimizing maintenance. Because you run it, plan for the operational work that comes with a private deployment.

How to decide

• Want to start now, minimal upkeep → hosted CoCalc.ai.
• Want your own machine or a single instance → local or single VM.
• Need a private deployment your organization runs → evaluate a customer-operated private deployment.

Across all three, the working model is the same: one durable project holding notebooks, terminals, code, documents, services, discussion, history, and recovery, with AI agents working in the same shared project as your files, software, and collaborators.

For current details on cost and licensing, see CoCalc's Pricing and licensing page, and use the support and sales channels for deployment-specific questions.

CoCalc is a collaborative computing platform built around persistent, shared Linux projects. Instead of scattering files, notebooks, terminals, and AI agent sessions across separate tools, it keeps them together in one durable project that you and your collaborators can return to over time. It is developed by SageMath, Inc., which has offered collaborative computing online since 2013. If your work involves running code, sharing notebooks, or coordinating with other people (or AI agents) on the same files, CoCalc is designed for that situation.

The core idea: one project, many workflows

The central unit in CoCalc is the project. A project is a persistent Linux environment that holds:

• Files and documents
• Jupyter-style notebooks
• Terminals
• Running services
• Discussion and collaboration history
• Version history and recovery

Because these live in the same place, you don't have to move data between a notebook, a terminal, and a chat tool. The project is the shared workspace, and everything happens inside it.

Where AI agents fit in

CoCalc's page evidence describes agents working "where your project lives." Concretely, that means:

• You can use the integrated Codex agent or Claude Code inside the project.
• You can also run other command-line agents in project terminals.
• Agents operate with the same files, tools, and running services your collaborators already use.
• Collaborators can see edits live, and you can restore earlier versions.

One important condition: Claude Code is an experimental preview on sites that enable it, and it works with your personal Claude Pro or Max subscription. So it isn't universally available, and it depends on your own subscription rather than being bundled by default.

The practical benefit is continuity. When an agent saves work, you can open it directly, or take over the task yourself in the same environment.

What you can actually do

How CoCalc can be run

CoCalc offers more than one deployment path, according to its site:

• Hosted CoCalc.ai — start on the hosted service.
• Local or single VM — run CoCalc locally or on one VM.
• Customer-operated private deployment — for evaluation of a private setup.

Pricing and licensing details are handled on a dedicated pricing page, so check there for current terms rather than assuming a free tier.

When CoCalc is a good fit

CoCalc makes sense if you:

• Want files, notebooks, terminals, and history in one persistent place
• Need collaborators to see changes as they happen
• Want AI agents to work directly on your project files rather than in an isolated sandbox
• Value version recovery as part of the workflow

It may be less suitable if you only need a one-off script run with no collaboration, no persistence, and no need to revisit the environment later.

To get started, the site points to CoCalc.ai, feature workflow pages, and documentation for deeper detail.

CoCalc publishes a dedicated Pricing and licensing page at cocalc.ai/pricing, and the site also points to support and sales channels for questions the page doesn't answer. Because CoCalc can be run in more than one way — hosted on CoCalc.ai, locally or on a single VM, or as a customer-operated private deployment — the right place to look depends on which path you're evaluating. If you plan to use the integrated Claude Code agent, note that it is an experimental preview on sites that enable it and works with your personal Claude Pro or Max subscription, so that cost sits outside CoCalc's own licensing.

Start with the pricing and licensing page

The official entry point is the Pricing and licensing link on cocalc.ai. Use it as the first stop for:

• What a hosted CoCalc.ai plan includes
• Licensing terms tied to how you run CoCalc
• Any plan tiers or options the site lists

The site does not publish specific prices, plan names, or payment methods in the material available here, so treat the pricing page itself as the source of truth rather than assuming a free tier or a particular billing model.

Match the pricing question to your deployment path

CoCalc describes three ways to run it, and cost and licensing questions differ for each:

If you already know you need a private or customer-operated deployment, go straight to support and sales — that path is described as something to "evaluate," which implies a conversation rather than a self-serve checkout.

Use support and sales for anything not on the page

The site lists support and sales as a separate channel from the pricing page. Reach for it when you need:

• Licensing clarification for a specific deployment model
• Volume, institutional, or private-deployment questions
• Confirmation of what a plan covers before committing

This is also the right channel if the pricing page leaves a gap between what's listed and what your team actually needs.

Factor in agent-related subscriptions

CoCalc's page states that the integrated Codex agent and Claude Code can both run in one project, alongside other command-line agents in project terminals. One licensing detail matters for budgeting:

• Claude Code is an experimental preview on sites that enable it.
• It works with your personal Claude Pro or Max subscription.

So if Claude Code is part of your evaluation, that subscription is a separate cost from CoCalc licensing, and availability depends on whether the site you use has the preview enabled. The Codex agent is described as integrated, without a stated separate subscription requirement in the available material.

A practical evaluation sequence

1. Open the Pricing and licensing page and note what applies to hosted CoCalc.ai.
2. Decide which deployment path fits — hosted, local/VM, or private — since that changes the licensing question.
3. If you're considering private or customer-operated deployment, contact support and sales before assuming self-serve terms.
4. If you'll use Claude Code, confirm the preview is enabled on your site and account for a personal Claude Pro or Max subscription separately.
5. For anything the page doesn't cover, use the support and sales channel rather than inferring terms.

CoCalc has been developed by SageMath, Inc. and describes itself as collaborative computing online since 2013, so the support and sales route is an established channel rather than a new contact point.