Developers
Stop leaving your editor to make a placeholder, an icon, an OG image, or a demo asset. Ask your agent, get the file in your project folder, keep shipping.
Loading…
Avocado CLI · v0.2
Generate from your editor
One command to authorize. One command to install. Then Claude Code, Cursor, or any MCP-aware agent talks to Avocado's 87+ image, video, music, and voice models, billed against your existing plan credits.
1
Install the CLI
One global install with npm. Node 18+.
$ npm install -g @avocado-ai-co/cli2
Authorize once
Opens your browser, you approve, a scoped key is saved to ~/.avocado/config.json.
$ avocado auth login3
Install into your agent
Wires the Avocado MCP server into Claude Code (or Cursor). Then just ask for an image.
$ avocado mcp install --client claude-codeTwo commands. Then your agent does the rest.
The CLI is the bridge: it authorizes your machine and wires Avocado into the editor you already live in. After that, generating an asset is just another thing you ask your agent to do.
From the terminal
Your terminal is now a render farm
Stay in the flow you're already in. Ask your agent for a hero shot, a 6-second video, or a voiceover and it calls Avocado on your behalf. No tab-switching to a web app, no copy-pasting prompts into a separate tool.
- 87+ image, video, music, and voice models, callable from chat
- Runs against your existing plan credits, no separate billing
- Free, open CLI. The models do the heavy lifting on Avocado's infra
Organized output
Assets land in folders, not buried in chat
Installing on Claude Code drops a skill that teaches the agent to save every generation to a local folder with descriptive filenames, a manifest.json, and captions, so a batch of product shots is ready to open in Finder or push to a CMS, not lost in a scroll-back.
- One folder per generation, zero-padded descriptive filenames
- manifest.json + captions.md written for every batch
- Long video jobs polled to completion before it reports done
In your editor
Lives inside Claude Code and Cursor
Anything that speaks the Model Context Protocol can use Avocado. The CLI installs the server into your editor's settings with a single flag, so generated images render right next to the code you're shipping them into.
- One flag: --client claude-code or --client cursor
- Generations appear inline in the chat panel
- Same context as your code, no app-switching tax
Every model
The whole library, one keystroke away
Nano Banana 2 for fast iteration, GPT-Image 2 when the text on the can has to be readable, Seedream 5 for photoreal at scale, Seedance 2.0 for video. Name the model, or let your agent pick the right one for the brief.
- 26+ image models, swappable mid-conversation
- Video, music, SFX, and voice from the same prompt window
- Whatever's on the web app is callable from the CLI
Edit and remix
Edit by describing the change, without leaving your editor
Point your agent at a reference image and tell it what to change. “Swap the background to a marble countertop.” “Make it the same brand shot in winter.” The CLI runs the edit on Avocado and hands the result back in the same chat, saved to the same folder.
- Nano Banana 2 for fast iteration, GPT-Image 2 when text matters
- Multi-image stitching, blending, reference-driven composition
- Each edit keeps the previous one in context. Layer changes without starting over
Built for people who live in a terminal
If your day already runs through Claude Code, Cursor, or a shell, the CLI meets you there instead of asking you to context-switch into yet another web app.
Indie hackers & creators
Ship the post and the product page without opening five apps. Hook image, vertical video, voiceover, all generated and organized from the same window you build in.
Agencies & creative teams
Script the boring parts. Batch-generate campaign variations from a prompt, organized into folders with manifests, ready to hand to the client or drop into the pipeline.
Why a CLI beats another browser tab
The whole point is to not break flow. Generation becomes part of the conversation you're already having with your coding agent.
1
Install once
Set it up in under a minute
Install, authorize in the browser, install into your agent. Three commands, then you never think about config files or API keys again.
Setup
npm i -g @avocado-ai-co/cli && avocado auth login && avocado mcp install --client claude-code
2
Prompt from chat
Ask your agent, not a separate tool
Describe the asset in plain language. The agent calls Avocado, polls long jobs to completion, and saves the result to a local folder with a manifest.
Example prompt
Using Avocado, generate 5 product photos of my matcha latte on a marble counter and save them to ./assets.
3
Ship it
Finish in the same window you started
The file is already on disk in your project. Drop it into the component, the deck, or the post. No downloading, no asset wrangling, no app-switching.
Example prompt
Now write three Instagram captions to go with the hero shot you just made.
Got any questions?
Common questions about installing and generating through the Avocado CLI.
Two commands away from generating in your editor.
Install
$ npm install -g @avocado-ai-co/cliAuthorize
$ avocado auth loginInstall into Claude Code
$ avocado mcp install --client claude-codePrefer Cursor? Swap the last flag for --client cursor.
New to Avocado? Start here.