Getting Started with AceKit for Coding Agents

Coding agents are great at editing files, but they often stop when a task needs the outside world: generating an image, searching the web, shortening a link, or calling a model-specific API. AceKit is a small installer that wires Ace Data Cloud capabilities into coding agents such as Claude Code, Codex CLI, Cursor, Gemini, and OpenCode. Instead of manually reading multiple API docs and writing one-off scripts, you install a skill toolkit once, provide a token, and let the agent follow each skill’s SKILL.md instructions. This guide walks through the practical setup and then looks at what happens behind the scenes.

What you can do

After installation, your agent can load skills for tasks that normally sit outside plain code editing. Examples include:

• generating a README cover image
• turning a script into a short video
• generating background music
• searching Google SERP in real time
• shortening a long link into a surl.id URL

The current skill list includes:

• acedatacloud-api
• ai-chatface-transform
• fish-audio
• flux-image
• google-search
• hailuo-video
• kling-video
• luma-video
• midjourney-image
• nano-banana-image
• producer-music
• seedance-video
• seedream-image
• short-url
• sora-video
• suno-music
• veo-video
• wan-video

That list can change over time, so use npx acekit list to inspect the current set on your machine.

How it works

AceKit is the front door. It detects supported coding agents, installs a set of agent skills, and uses one Ace Data Cloud token for authentication. The requirements and fields are straightforward:

• Node.js 18+
• install command: npx acekit
• optional environment variable: ACEDATACLOUD_API_TOKEN
• Claude Code skill path: ~/.claude/skills
• Codex CLI skill path: ~/.agents/skills
• Cursor / Gemini / OpenCode skill path: ~/.agents/skills

Each installed skill includes a written SKILL.md manual. That matters because the agent does not need to guess how a capability works. The skill can describe authentication, parameters, polling, retries, and how to return the final artifact.

Step 1: Prepare the token

AceKit can prompt for a token interactively on first run. For repeatable setup, export it before installing:

export ACEDATACLOUD_API_TOKEN=your_token

Do not commit real tokens to a repository. If you want teammates to reproduce the setup, document the variable name and let each developer provide their own value locally.

Step 2: Install the skill toolkit

Run the installer:

npx acekit

A documented run on a machine with Claude Code installed looks like this:

🃏 AceKit - wiring AI into your coding agent
✓ Detected: Claude Code
📦 Installing the AceData skill toolkit via @acedatacloud/skills …
Installed 19 skills to ~/.claude/skills
✅ Done. Try it now - ask your agent:
"generate a hero image for this README"
"turn this script into a 30-second video"

After installation, reload the agent so it can discover the new skills.

Step 3: Try a small synchronous task

Start with something deterministic, like shortening a URL. In plain language, you can ask the agent:

Use the short-url skill to shorten https://platform.acedata.cloud/services

Behind the scenes, the direct API shape is:

curl -X POST https://api.acedata.cloud/shorturl \
  -H "Authorization: Bearer $KEY" \
  -H "Content-Type: application/json" \
  -d '{"content":"https://platform.acedata.cloud/services"}'

This is a useful smoke test because it is synchronous and quick. If the agent can call this skill and return a link, your token and skill installation are working.

Step 4: Add real-time search to agent work

Another simple but high-value test is search. The example call is:

curl -X POST https://api.acedata.cloud/serp/google \
  -H "Authorization: Bearer $KEY" \
  -H "Content-Type: application/json" \
  -d '{"query":"OpenAI Sora release date","type":"search","number":3}'

In an agent workflow, the prompt can be more natural: ask it to search for a current API behavior, compare two libraries, or summarize recent documentation. The key improvement is that the agent can fetch fresh results instead of relying only on model memory.

Step 5: Understand asynchronous tools

Some capabilities are not instant. The image example calls the Nano Banana image endpoint:

curl -X POST https://api.acedata.cloud/nano-banana/images \
  -H "Authorization: Bearer $KEY" \
  -H "Content-Type: application/json" \
  -d '{"action":"generate","model":"nano-banana", "prompt":"a cute cartoon banana mascot waving hello, flat vector logo, white background"}'

The important lesson is not the mascot prompt. It is that the installed skill can handle the workflow around the API: building the request, authenticating, polling when needed, and returning the final image to the conversation. That is where agent skills become more useful than a loose collection of curl snippets. They give the agent operational instructions, not just endpoint names.

Step 6: Know where skills are installed

The installation locations are:

• Claude Code: ~/.claude/skills
• Codex CLI: ~/.agents/skills
• Cursor / Gemini / OpenCode: ~/.agents/skills

If something does not appear in your agent, check the expected directory first, then restart the agent. For multi-agent setups, remember that some tools may read from ~/.agents/skills while Claude Code reads from ~/.claude/skills.

Optional: Claude Code environment settings

Ace Data Cloud can be used with Anthropic-compatible /v1/messages settings for Claude Code. The local project file is .claude/settings.local.json:

{
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "your token",
    "ANTHROPIC_BASE_URL": "https://api.acedata.cloud"
  }
}

Keep this file local if it contains credentials.

Final notes

AceKit is most useful when your coding agent already understands the codebase but needs safe access to external capabilities: search, media generation, link tools, or model calls. Start with a small synchronous skill, verify the install path, then try longer-running tasks where polling matters. The full setup details are in the AceKit Toolkit guide.