# Create AI Agent by Describing It | FlyMy.AI

> Turn a plain-English job into a cloud agent, review a real run, and freeze the accepted workflow for reuse.

Page: https://flymy.ai/use-cases/create-ai-agent-by-describing-it/

Describe the job. Review the result. Reuse the workflow.

# Create AI agent by describing it

Describe the job, the inputs and the result you want. FlyMy.AI can turn that brief into a cloud agent with selected tools. Connect the required services, inspect one completed run, then freeze the accepted workflow. Use the same workflow from your application, a schedule or a connected coding assistant.

[Build this agent](https://app.flymy.ai/signup) [Read the recipe](https://flymy.ai/use-cases/create-ai-agent-by-describing-it/#recipe)

## From a plain-English brief to a reusable agent

1. ### Write an acceptance example Specify the input, allowed sources, output format and destination. For example: research three named companies and return a source-linked summary. Avoid an open-ended instruction to search everything.
2. ### Connect the assistant and required tools Follow the FlyMy.AI MCP setup from Claude Code or Cursor. Discover the exact tools and connect your accounts. The Python quickstart below shows the equivalent authoring API, with explicit tool IDs.
3. ### Run once and inspect the evidence Create the agent, run a small example and inspect its status, result and tool activity. A created agent is not proof that its workflow succeeded. Change the goal or connections if the result does not meet your acceptance example.
4. ### Freeze the accepted run Freeze the successful execution, wait for compilation and test the frozen workflow. Keep the compilation ID for later app calls or scheduling; do not create another agent for every input row.

## Deploy AI agent from Claude Code

Connect FlyMy.AI MCP, provide the same brief and ask the assistant to create, test and freeze the agent. The cloud workflow runs on FlyMy.AI. Claude Code and Cursor are authoring clients; their local process does not need to stay open for a configured schedule.

Python authoring example copied from the quickstart - [source documentation](https://docs.flymy.ai/agents/quickstart)

```
from flymyai import AgentClient

client = AgentClient(api_key="fly-***")

# Register tools first - each returns a Tool object with an integer .id
web_search = client.tools.create(mcp_tool="tavily")
browser = client.tools.create(mcp_tool="browser")

agent = client.agents.create(
    name="Web Researcher",
    goal="Search the web for the latest breakthroughs in quantum computing and return a concise summary with sources.",
    tools=[web_search.id, browser.id]
)

print(f"Agent created: {agent.id}")
```

Install flymyai first. Replace the placeholder key in your local server environment. The example goal is deliberately unchanged from the docs; adapt the brief and tool selection to your job.

## What you get

Illustrative output - example data

```
job: Research three companies
status: completed
result: A concise summary with source URLs
next: Review the result, then freeze this execution
```

### Where the result goes

The result is available on the run. Read it through the documented SDK or API. Sending it to another service requires that service connection and an explicit delivery step in the workflow.

## Cost and operating limits

Agent tokens and paid tool/model calls are metered. Keep the acceptance run small, review its cost and set a budget before repeated execution. Creating a reusable brief does not make future executions free.

[Check current pricing](https://flymy.ai/pricing/)

## Choose the right approach

**Implementation choices for this job**

| Approach | What to evaluate |
|---|---|
| FlyMy.AI cloud agent | Use when the accepted workflow needs connected tools and later scheduling or app execution. |
| A local script | Use when you already have deterministic application code and will operate its runtime, credentials and scheduling yourself. |
| A coding-assistant conversation | Useful for authoring and interactive review. Confirm the separate deployed workflow before relying on unattended execution. |

## Frequently asked questions

**Do I need to write the agent code?**

A plain-English goal can define the job. You still need to connect tools, provide valid inputs and review a successful execution. The SDK is another authoring option.

**Can I keep using the same agent?**

Yes. Use input variables and freeze an accepted run for repeated execution. Review and test changes before selecting a new live version.

**Does connecting MCP publish my private data?**

No public publication is part of this recipe. Use only the connections and sources you intend the agent to access.

## Related jobs

Use case

### [Scheduled AI agents](https://flymy.ai/use-cases/scheduled-agents/)

Schedule a tested agent for a morning report, recurring research or structured output using cron and an explicit timezone.

[Read the guide →](https://flymy.ai/use-cases/scheduled-agents/)

Use case

### [AI agents for SaaS products](https://flymy.ai/use-cases/ai-agents-for-saas/)

Embed a tested agent behind your SaaS backend, with customer-specific service connections and explicit builder billing.

[Read the guide →](https://flymy.ai/use-cases/ai-agents-for-saas/)

## Sources

[Agent quickstart](https://docs.flymy.ai/agents/quickstart), [Operating guide](https://flymy.ai/skill.md), [MCP setup](https://flymy.ai/mcp.md), [Pricing](https://flymy.ai/pricing.md)

Sources checked 2026-10-23. Examples are documentation excerpts, not a claim that a paid run has been performed.

---

For agents: the FlyMyAI operating guide (sign up, connect, run models, tools and agents) is https://flymy.ai/skill.md, and https://flymy.ai/llms.txt lists the other pages and the documentation.