Model discovery
Find available models and identify capabilities and versions that fit your task.
AGENT TOOLS / APIMART
Give Codex, Claude, Cursor and other agents a way to discover models, understand parameters and run image and video tasks.
“Find a model for a product hero image, generate a perfume ad, and tell me the task result.”
Identify the image task
Compare models and inputs
Call the APIMart API
Get status and files
Example of a completed task
The agent prepares inputs from model docs, submits a task and retrieves results. This preview sends no requests and uses no credits.Select a prompt to explore an example workflow. This simulation sends no requests and consumes no credits.
Use APIMart MCP list_models to list 5 available models. Do not generate content.
→ APIMart / list_models
{"limit": 5}
Example display names, not API model IDs:
01 GPT Image 2.5
02 Nano Banana Pro
03 Seedream 5.0 Pro
04 Z-Image Turbo
05 Midjourney
✓ Example data. Actual counts and model IDs depend on the results returned for your key.CHOOSE A PATH
Choose docs, remote tools or a local Skill. Supported operations depend on the API, tools and client you use.
SETUP GUIDE
Choose a method and follow the steps. Installation commands and test prompts are ready to copy.
APIMart Agent Tools
macOS · zsh · Codex desktop · Production
MCP / Step 01
Create a valid APIMart API key, then run this command in zsh on macOS. Enter the key when prompted; input is hidden.
Get API KeyTerminal command
read -s "APIMART_API_KEY?APIMart API Key: "
echo
launchctl setenv APIMART_API_KEY "$APIMART_API_KEY"
unset APIMART_API_KEYPaste the actual key only at the terminal input prompt. Keep it out of commands, chats and screenshots. Revoke and replace any exposed key.
This configuration applies to desktop apps restarted during your current login session. Fully quit and reopen Codex desktop. For the terminal CLI, read the key with hidden input, run export APIMART_API_KEY, then start Codex from that same terminal; omit the unset command shown here. The key format below is only an illustration; do not run or replace it.
sk-••••••••••••••••COMPARISON
Use AI docs for quick questions, MCP for everyday tool calls, or Skill for local scripts and repeatable workflows.
| Method | Best for | Setup | Execution | Main benefit | Keep in mind | Get started |
|---|---|---|---|---|---|---|
| AI Docsllms.txt | Quick questions and API code | No install; copy a prompt | Your agent reads docs and runs code | Lightweight and flexible | Web-reading access required; API calls still need a key | |
| MCPRecommended | Frequent calls and daily creation | Remote service URL and API key | APIMart remote MCP | Structured tools; no local MCP package | Requires network and client support; limited to available tools | |
| Agent SkillLocal package | Repeatable and team workflows | Install the full package from GitHub | Local scripts call the API | Reusable workflow, independent of MCP | Requires Node.js; updates may require reinstalling |
CAPABILITIES
Discover models, understand documentation and handle common image and video tasks. Confirm that the model, tools or local client support the operation you need.
Find available models and identify capabilities and versions that fit your task.
Read input parameters, usage instructions and limitations before making a call.
Select a model, prepare parameters, submit a task, check its status and retrieve results.
Call video models and track task progress, execution status and final results.
SECURITY FIRST
Pass credentials through environment variables to reduce accidental exposure in terminal history, screenshots and conversations.
AI docs help your agent understand the API; reading them does not trigger model calls.
MCP and Skill read your key from an environment variable. Keep it out of prompts and installation commands.
Actual generation is billed to your APIMart account. Verify your setup with a model query first.
FAQ
MCP and Skill work independently. AI docs need no installation—just copy a prompt.