Skip to main content
You have probably encountered this situation too: you are using Claude Code, Cursor, or Codex, writing code smoothly, and suddenly need an illustration, a piece of background music, a short video, or want it to look up the latest information along the way. Then you discover—agents themselves do not know how to do these things. You have to find APIs yourself, read documentation, configure keys, write calling code, handle asynchronous polling; after all that hassle, your inspiration has long gone cold. AceKit is here to save you from this entire set of hassles. With one command, it installs image generation, video generation, music generation, web search, LLM conversations, and other capabilities from AceData Cloud into your coding agent as a complete package. After installation, you can just speak naturally; the agent will load the corresponding skills itself, authenticate itself, poll itself, and finish the work.

Understand What It Does in 30 Seconds

Before installation vs. after installation: In one sentence: your coding agent goes from “only able to write code” to “also able to generate images, generate videos, generate music, and search the web.”

Get Started in Three Steps

The entire process can be completed in a few minutes and requires Node.js 18+. Step 1 · Get an API Token. Go to the console and copy a Token—this single one is shared by all capabilities. Your first application comes with free credits, so you can try it for free first. Copy the API Token in the console Step 2 · Install. One command:
It automatically detects the coding agents on your machine and installs the complete skill set. A real run looks like this (this machine only has Claude Code installed):
If no Token is set, the first run will interactively prompt you to paste one; you can also run export ACEDATACLOUD_API_TOKEN=yourtoken in advance for a completely unattended installation. The number of skills will change as platform capabilities increase; refer to the current list displayed by npx acekit list.
Step 3 · Speak naturally. Reload your agent, then directly give it instructions, for example:
The agent will load the skill itself, initiate the call itself, and give you a real, accessible result.

Run It for Real (With Output)

The following four examples are all real calls with results returned on the spot, not demonstrations. The $KEY in the commands is your API Token. ① Shorten a URL (synchronous, returns in seconds)
Returns https://surl.id/r/1uKp2kilXQ—opening it will redirect with 302 back to the original services page. ② Web search (synchronous)
Gets real-time Google results (excerpt):
③ LLM conversation (synchronous, one API for 50+ models)
④ Image generation (asynchronous, about 20 seconds, the agent automatically polls until the image is generated)
After about 22 seconds, it returns a real image (HTTP 200 · image/png · 317 KB · approximately 0.13 credit this time): An image actually generated by AceKit You do not actually need to remember these APIs—after installing the skills, the agent will assemble the correct requests itself according to each SKILL.md, and handle authentication and polling. The above only lays out for you “what it does behind the scenes.”

What Exactly Does It Install

AceKit is the “front door” to AceData’s entire developer tool surface. At present, a single installation installs the skill toolkit (Skills):
  • A set of Agent skills: each skill is a pre-written SKILL.md operation manual—authentication, asynchronous polling, error retries, and how to fill in parameters are all clearly documented; the agent only needs to follow it. Run npx acekit list to see the current list (there are 19 at the time of writing, and it continues to grow):
  • One unified Token: One ACEDATACLOUD_API_TOKEN unlocks all capabilities, billed based on actual usage.
Classified by capability (subject to npx acekit list):
For these same capabilities, AceData also provides standalone MCP servers (Suno, Midjourney, Sora, SERP… published to PyPI / VS Code / Smithery / Docker), command-line CLI, and Python / TypeScript SDK, which can be used separately as needed. AceKit v0.2 plans to support single-capability installation via acekit add <name>, as well as automatically writing MCP configurations for each agent.

Which Agents Are Supported

During installation, AceKit automatically detects the agents installed on the machine and installs only into the detected ones.

Easter Egg: Use the Same Token to Power Claude Code Itself

AceData’s interface is also compatible with Anthropic’s /v1/messages format, so you do not need an official Anthropic subscription. You can use the same AceData Token to power Claude Code. Create .claude/settings.local.json in the project root directory:
Simply restart Claude Code—it can use the full Claude model lineup while also gaining the “hands” to generate images, videos, and music through AceKit. For more details, see the Claude Code Integration Guide.

Start for Free

AceKit is open source (MIT), with the repository at github.com/AceDataCloud/AceKit. If you find it useful, please give it a ⭐.