1
Check your environment and choose a client
- The installer requires Node.js 18 or newer.
- Using PoYo CLI requires Node.js 20 or newer.
Coding client
Official PoYo Agent SkillOpen source · Verifiable
Choose the right model. Finish the task safely.
Install poyo-ai-models in your coding agent, connect PoYo through MCP or CLI, and give the agent a reliable path from live model discovery to final delivery.
$npx skills add PoyoAPI/poyo-devtools --skill poyo-ai-modelsBuilt for coding agents such as Codex, Claude Code, Cursor, and GitHub Copilot. Not using a coding agent? Open the PoYo creation platform
Complete the basic setup in four steps
Choose your client and install scope, configure the API key, then run a read-only verification.
2
Install the official Skill
Choose your coding client and let the guide generate the exact project or global command.
Choose where to install
Choose a client first to generate a safe, deterministic command.After installation, reload or restart the client so it can discover the Skill.
3
Configure a PoYo API key
Use an environment variable or your client's secret manager. Never put the key in a prompt or repository.
Open API KeysConfigure your API key
$export POYO_API_KEY="your_api_key"This command only affects the current terminal session. Use your client's secret manager for a persistent setup, and never commit the key.
4
Verify without a paid task
Ask the agent to confirm the Skill and execution path, then query only models, input requirements, and public pricing.
Confirm that the poyo-ai-models Skill is loaded and state whether you will execute through PoYo hosted MCP or PoYo CLI. Only query currently available image models, input requirements, and public pricing. Do not submit a paid task.What the agent should confirm
- poyo-ai-models is loaded and available.
- One execution path is ready: PoYo MCP or PoYo CLI.
- The agent only returns live models, input requirements, and public pricing—without submitting a paid task.
Tasks you can hand directly to your agent
Every example asks the agent to compare current capabilities and pricing before submitting against the real schema.
Agent request / video
Turn a product image into a short video
“Turn this product image into a five-second ecommerce showcase with a slow push-in and no subject distortion. Compare image-to-video models and pricing before execution.”
Discover image-to-video models → read duration limits → execute → poll task_id
What the Skill does
One compact operating guide covers the four decisions that most often cause model tasks to fail.
Discover available models
Search the current catalog by task intent instead of inventing model IDs from memory.
catalog search
Read parameters and pricing
Inspect live schemas, required fields, and public pricing before execution.
schema + pricing
Choose an execution path
Use MCP for agent interaction and the PoYo CLI for scripts and CI jobs.
MCP / CLI
Resume long-running work
Save task_id and distinguish waiting, success, and failure while polling.
check job
A complete task lifecycle, not a list of prompts
The agent moves from discovery to live schema checks, selects an execution path, preserves task state, and reports usable results.
The Skill never replaces PoYo's live catalog, pricing, or task status. It teaches the agent when and how to read them.
agent / poyo-ai-modelsready
01Search modelsFind candidate capabilities in the live Capability Catalog.
02Read the factsCheck schemas, limits, public pricing, and account status.
03Choose the pathUse MCP or CLI based on interaction and automation needs.
04Execute safelySubmit valid parameters and preserve idempotency when needed.
05Resume long jobsSave task_id and continue status checks at sensible intervals.
06Deliver the resultReport model, status, charge information, and final output.
Never call waiting complete
The agent must accurately distinguish pending, running, succeeded, and failed.
Never retry blindly
Check balance and status, and preserve idempotency data and task_id to reduce duplicate charges.
When should you use MCP, CLI, or REST?
After installation, agents still choose MCP or CLI for the environment. Your application services continue to use the REST API directly.
PoYo MCP
Agent conversations and coding clients
Reasoning, discovery, and execution happen together
Explore MCPPoYo CLI
Terminals, scripts, and CI
Reproducible commands, batches, and automation
View GitHubREST API
Product backends and stable application traffic
Direct integration, queue control, or token streaming
Read docsCommon questions
The most common questions, answered with the real product boundaries.
Still have a question? Contact PoYo support→
It is a set of execution instructions for AI agents, covering model discovery, Schema and pricing checks, MCP/CLI selection, long-job recovery, and result reporting. It is not a new model API and does not train anything on your machine.
MCP gives the agent tools; the Skill tells it in which order and under which conditions to use them. MCP alone can execute tasks, but the Skill reduces invented parameters, duplicate submissions, and long-job status mistakes.
Protected account lookups and real execution require an API key. Store it in an environment variable or client secret manager, never in prompts, project files, or chat history.
Yes. Choose PowerShell or Command Prompt in the API key step. The installer requires Node.js 18+, while PoYo CLI requires Node.js 20+.
Reload the client, confirm that poyo-ai-models is installed in the intended project or globally, then verify that either the existing PoYo MCP connection or PoYo CLI is available. Run the read-only verification prompt again before submitting a paid task.
No. The Skill does not change model pricing. Execution is charged according to PoYo's current public pricing and the actual request; the agent should query and explain the basis before running it.
Install the PoYo Skill and run your first task
Create an API key, install the official Skill, and describe the outcome. One verifiable workflow handles model selection, parameter checks, and long-job recovery.