Claude Code Image Generation: Kollab CLI Setup

Short answer

To add image generation to Claude Code or Cursor, install Kollab CLI and sign in with kollab login. Send a request with kollab task ask from your terminal or through your agent. Kollab runs the task in its cloud, so you do not supply model-provider API keys.[1]

View CLI setup
A lone developer at a desk on a mountain ledge sends a thread of golden light across a misty canyon to a cloud-top studio-forge, where picture panels, a film strip, a 3D vase and a sound-wave orb rise into the sky.
Type
Guide
Published
October 26, 2026
Last checked
October 26, 2026
Sources
8

Use Kollab CLI from the terminal in Claude Code or Cursor to send media requests to your Kollab workspace. The commands below cover installation, sign-in and a first image request.[1]

Claude Code and Cursor support Skills for loading workflow instructions. For media generation, give your agent the CLI command shown below.[2][3]

What does the Skill add to your agent?

Kollab's CLI page describes the Skill as instructions for using the CLI. The CLI performs the actions. Claude Code and Cursor each provide a way to load those instructions.[1]

In Claude Code, Skills can run when relevant or through /skill-name. In Cursor, type / in Agent chat and find the installed Skill. Cursor discovers Skills when it starts, so restart Cursor if the installed Skill does not appear.[2][3]

Claude Code picks up changes in existing Skill directories during the current session. If the top-level directory did not exist when the session began, run /reload-skills.[2]

Do you need to supply API keys?

Kollab's CLI page says, “You do not need to provide API keys.” Kollab configures the model services for its cloud tasks. You still sign in to Kollab. Checked October 26, 2026.[1]

Messages, images, videos and automated tasks draw from the same workspace credit balance. Heavier models and longer runs cost more. Review Kollab's pricing terms before running a request.[4]

How do you install, sign in and make a first request?

Install the CLI, add the Skill to your agent, sign in and send an image request.[1][5]

  1. 1

    Install the CLI in your terminal

    Check its version and available commands.

    npm i -g kollab
kollab --version
kollab --help

    The install command comes from the official CLI page; --help lists the available commands.[1][5]

  2. 2

    Install the agent Skill

    The installer asks which coding agents to install to. Choose Claude Code or Cursor. npx skills requires Node.js 22.20 or later.[6]

    npx skills add KollabTeam/Kollab@kollab-cli

    Skills install into the current project by default. Add -g to make the Skill available across projects.[6] Kollab publishes the installation command.[1][2][3]

  3. 3

    Sign in to Kollab

    The CLI opens a browser and asks you to choose a workspace.

    kollab login

    kollab login --help also lists --token for sign-in without a browser. Treat the token as a secret.[1][5]

  4. 4

    Request your first image

    Run this command in the terminal, or tell your agent to run it.

    kollab task ask --model-tier auto --question "Create an image of a small ceramic planter on a plain background. Return the generated result."

    kollab task ask --help lists --question and --model-tier auto. The command returns JSON with stream events, persisted messages and artifacts. Find the saved image under data.persisted.artifacts and its preview link in the persisted answer.[5]

How do you request video, 3D and search?

Use the same task command with a different request. Kollab lists these capability categories on its CLI page. Its public 3D description covers text or a reference image turned into a .glb file through Tripo3D.[1]

After signing in, run kollab tool list to see managed capabilities and their parameters. The following requests use the same task ask command as the image example.[5]

Video
kollab task ask --model-tier auto --question "Create a short video of a ceramic planter turning slowly on a plain background. Return the generated result."
3D
kollab task ask --model-tier auto --question "Generate a 3D model of a small ceramic planter and return the .glb result."
Search
kollab task ask --model-tier auto --question "Search the web for recent ceramic planter design references and include source links."

For a direct tool call, kollab tool run <tool> accepts --input-json, --input-file or standard input. Use the parameters returned by tool list. Set --output-dir to choose where media files are downloaded.[5]

How should you compare MCP and CLI for AI agents?

A Skill supplies instructions that the agent loads when needed. A CLI is an executable the agent runs in the terminal. An MCP integration exposes external tools through a registered server. These are the practical differences when comparing MCP vs CLI for AI agents.[1][2][3][7]

Skill, Kollab CLI and MCP compared

Skill, Kollab CLI and MCP compared
ComparisonWhat the agent loadsHow it calls the serviceSign-in
SkillWorkflow instructions, loaded when needed.[2][3]Instructions tell the agent which commands to run.[1]Follow the service's sign-in flow.[1]
Kollab CLIAn installed executable.[1]Terminal commands send tasks to Kollab's cloud.[1]kollab login selects a workspace.[5]
MCP integrationTools exposed by a registered server.[7]Artificial Studio's tutorial registers a remote service with claude mcp add --transport http.[7]That tutorial uses browser sign-in on first use.[7]

Kollab also lists a managed-tool server over stdio in kollab mcp serve --help. This guide uses the CLI route; MCP setup and a performance comparison are outside its scope.

What if your team starts in ChatGPT?

This guide covers Claude Code and Cursor. It does not cover a direct ChatGPT integration. To use the steps above, give the finished prompt to the agent that runs Kollab CLI.

Claude Agent Skills Explained

Anthropic · Claude Agent Skills Explained · [8]Watch on YouTube(Opens in a new tab)

Video context and disclosure

This video gives background on the Skill layer and how it relates to MCP; its description lists an Agent Skills vs MCP Servers chapter at 01:42. It is an Anthropic product explainer, not a Kollab demonstration or independent review.[8]

Public metadata and embed permission were checked on October 26, 2026; full playback has not been reviewed.

Frequently asked questions

Sources

  1. 1.

    KollabKollab CLI: Operate Your Kollab Workspace From the Command Line (Opens in a new tab)kollab.im. .

  2. 2.

    AnthropicExtend Claude with skills (Opens in a new tab)code.claude.com. .

  3. 3.

    CursorAgent Skills (Opens in a new tab)cursor.com. .

  4. 4.

    KollabKollab Pricing for AI Agent Workspaces (Opens in a new tab)kollab.im. .

  5. 5.

    Kollab CLI package publisherkollab (Opens in a new tab)www.npmjs.com. .

  6. 6.

    Vercel Labsskills CLI README (Opens in a new tab)github.com. .

  7. 7.

    Artificial StudioHow to generate 3D models in Claude and import them into Blender (Opens in a new tab)app.artificialstudio.ai. .

  8. 8.

    AnthropicClaude Agent Skills Explained (Opens in a new tab)YouTube. .

Related guides

Prepared by Kollab from the sources below with AI-assisted drafting, then reviewed by another AI model against those sources. This page brings together the public sources listed below. Last checked October 26, 2026. Command syntax matches Kollab CLI 0.1.25, the current version on the official npm registry.

Last checked: