CoCalc documentation

Move legacy cocalc.com billing credit and projects into CoCalc.ai using the same verified email address.

Create a durable Linux workspace for files, notebooks, terminals, chat, and agents.

Create and check a small analysis in Essential, save its context, and let a collaborator or viewer pick it up.

Store runtime credentials as encrypted, read-only files, select secrets to copy with the CLI, and check mount-refresh results.

Use durable collaborative terminals backed by real project Linux processes.

Find, open, create, and organize the CoCalc projects you can access.

Publish an exact file, a folder, or the whole project with an unlisted signed-in share link.

Use task files for shared checklists, project plans, and durable TODO lists.

Choose, customize, and reuse the Linux software stack for a project.

Create a CoCalc AI project with the RStudio and Jupyter image, then launch RStudio Server.

Publish project images with metadata, public slugs, discovery actions, app launchers, and CLI automation.

Invite collaborators or read-only reviewers, select visible files, and manage access as the work changes.

Start an agent with a request, review its results, and continue the work from one workspace.

A systematic feature reference for Codex, with Claude Code support levels, limitations, and links to detailed guides.

Give an agent temporary access to selected CoCalc projects or all your projects.

Connect ChatGPT or OpenAI API access for Codex and project code.

Use Codex inside a durable project workspace with files, terminals, and notebooks.

Use Claude Code in CoCalc: experimental setup, images, Agent Networks, credentials, billing, and the security model.

Choose access, models, reasoning, speed, defaults, and parallel workers.

Steer running work, queue follow-ups, send to existing threads from the CLI, choose a working directory, and fork context.

Manage continuing objectives and answer blocking or asynchronous questions.

Schedule Codex prompts or Bash commands, review runs, and pause automation.

Choose completion notifications and inspect or stop ongoing Codex sessions.

Supply document context, select a conversation, review prepared prompts, and reuse generated images.

Run a complete example twice, compare input hashes and saved results, and record the research environment.

Read a synthetic CSV one row at a time, check saved means and variances independently, and recognize incomplete results.

Transfer a notebook and its data, select the right kernel, replace platform-specific assumptions, and check a fresh run.

Run a checkpointed Python example, deliberately fail it, resume unfinished cases, and verify the saved results.

Compare one and two worker processes, check numerical integrals independently, and verify complete saved results.

Compile a C numerical routine, call it from Python, compare independent results, and retain source and build evidence.

Upload inputs with the CLI, execute in a CoCalc project, inspect the remote result, and retrieve a checked artifact.

Choose file history, snapshots, backups, or a clone, and practice restoring a checked copy without overwriting current work.

Launch a small managed HTTP app, verify readiness and collaborator access, inspect failures, and stop it.

Start a bounded CLI agent task, retain its continuation identifier, inspect sessions, and interrupt a selected task.

Check compute and CUDA software separately, calculate on a GPU, and save device and result evidence.

Render a complete R analysis and plot to HTML, inspect the output, and preserve the inputs and environment for a collaborator.

Use the CoCalc CLI for authenticated docs, browser, notebook, and project automation.

Install the CLI, sign in on your own computer, and inspect a project step by step.

Choose profiles, environment credentials, project context, browser targets, and fresh authentication.

Navigate command families for files, notebooks, collaborative documents, agents, browsers, and operations.

Parse JSON, check remote exit codes, and recover operations and asynchronous execution jobs.

Read, check, edit, and save live text with version and hash expectations.

Insert and run live notebook cells, inspect outputs, and recover detached runs.

Resolve browser targets, inspect the supported API, verify UI actions, and recover asynchronous work.

Create a disabled daily task, verify its configuration, and manage runs and updates.

Create and update workspace records, leave durable notices, and distinguish messages from agent turns.

Run complete document builds and distinguish CLI, server, browser, and agent-skill versions.

Use the limited CoCalc HTTP API carefully, and prefer cocalc-cli for most automation.

Use persistent collaborative Linux shell sessions inside CoCalc projects.

Launch Wayland and X11 applications, switch their windows, and connect terminals or notebooks to the graphical session.

Connect to cocalc.ai projects from a computer or another CoCalc project using SSH.

Use the project filesystem as the shared place for notebooks, scripts, datasets, and output.

Create, open, upload, rename, move, and organize project files.

Write README files, notes, instructions, math, code blocks, and collaborative documentation.

Create presentation slides that live with the project files they explain.

Sketch diagrams, lecture notes, and visual plans in a collaborative project file.

Inspect, compare, and recover the history of files in a project.

Use Git for repository history alongside TimeTravel for file-focused recovery.

Create collaborative notebooks with kernels and output capture in the project backend.

Use collaborative durable Jupyter notebooks, including legacy Sage worksheet conversion and rerun checks.

Navigate, run, and present a notebook in the content-first Studio view, with markdown headings as sections.

Use another machine's Jupyter kernel or GPU from a collaborative CoCalc notebook, without automatically synchronizing files.

Create a custom Jupyter kernel backed by a uv-managed Python virtual environment.

Install GNU Octave and octave-kernel into an existing CoCalc AI project without changing the default Python kernel.

Recover from Jupyter kernels that crash, restart, or fail to start.

Identify maintenance, resource-pressure, disk-quota, and host-connection start blocks and choose the next check.

Diagnose low-memory warnings, out-of-memory kills, and notebook kernel restarts.

Diagnose sign-in, websocket, stale browser state, and network connection problems.

Use real Python through notebooks, scripts, terminals, virtual environments, and papers.

Write and build LaTeX papers, assignments, reports, figures, and bibliographies.

Write reproducible R reports with Markdown prose, R chunks, plots, and rendered output.

Install CoCalc Star on a public Ubuntu VM, complete first-admin setup, and identify connections and retained state before maintenance.

Install CoCalc Star inside a local Ubuntu VM for private, fast, offline-friendly work on your own computer.

Install real apt-managed Chromium on Ubuntu from ppa:xtradeb/apps and prevent Ubuntu's snap transition package from returning.

Temporarily let a trusted CoCalc project SSH back to your computer through a reverse SSH tunnel.

Compare project hosts, managed VMs, and remote kernels, then check machine fit, availability, permissions, and readiness.

Run projects on dedicated or cloud-backed hosts and understand CPU sharing, RAM limits, and GPU access.

Delegate host access and understand shared-pool tiers, private-host RAM defaults, and per-project caps.

Move projects between hosts while accounting for backups, snapshots, region changes, and SSH.

Understand start, stop, restart, drain, deprovision, and delete actions for project hosts.

Explain spot retry windows, standard fallback, probes, and returning from fallback to spot.

Know which host edits are online, which require restart, and which require deprovision.

Read host reliability, availability, outage exposure, planned downtime, and day-grid signals.

Understand runtime software, managed daemons, reconcile, upgrades, drift, and rollbacks.

Understand host disk capacity, storage mode, backups, snapshots, and online disk growth.

Use host-scoped /scratch storage without confusing it with project storage, backups, or moves.

Interpret host metrics and logs to investigate resource pressure, provisioning, and runtime failures.

Set up, rehearse, run, and end an in-person exam on a private project host, with a clean, network-isolated notebook project for each student.

Discuss project work with collaborators and AI assistants in durable chat files.

Notify collaborators with @mentions and return to the relevant project context.

Run computational courses with student projects, assignments, collection, grading, and feedback.

Understand course-funded student VMs, spending limits, and what happens when funding or storage retention ends.

Configure student pay, instructor-paid seats, site licenses, course start dates, and grace periods.

Explain each student-project restriction option and what it really disables.

Use a common writable project shared by all students, instructors, and TAs.

Choose and roll out managed images for course projects.

Assign, collect, grade, and return computational work in student projects.

Use nbgrader for structured Jupyter notebook grading in CoCalc courses.

Jump between projects, files, editor frames, and settings using the keyboard.

Search version-matched CoCalc-ai docs from the public site or inside a project.

Use stable docs action ids to open the right UI from docs or Codex.

Use scoped browser-session automation to inspect UI and verify docs.