> ## Documentation Index
> Fetch the complete documentation index at: https://cadenya.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Create and publish an agent
> Create a draft agent in the dashboard, configure its first variation, publish it, and send it an objective.
An agent is the stable identity your application runs. A variation supplies the prompt, model, constraints, and assignments that determine how it runs.
This guide builds a small `Concise Helper` agent with one variation and no tools. The dashboard handles configuration. Code appears only when you send the finished agent an objective.
## What you need
* A Cadenya workspace where you can manage agents.
* At least one enabled model in the workspace.
* For the final step, an API key with **Objectives: Manage** and the workspace ID.
## Create the draft agent
Select **Agents**, click **Create Agent**, and enter:
| Field | Value | Why it matters |
| --------------- | ------------------------------------------------------ | ------------------------------------------------------------------------------------------- |
| **Name** | `Concise Helper` | The dashboard label for the agent |
| **External ID** | `concise-helper` | A stable identifier your application can keep across environments |
| **Description** | `Explains Cadenya concepts in two or three sentences.` | Describes the agent's purpose and becomes important if another agent uses it as a sub-agent |
Leave the advanced sections at their defaults and click **Create agent**. The new agent opens in **Draft** state with **No variations configured**.
The create form does not configure a prompt, model, or publication state. Those belong to the next steps. A draft agent cannot accept objectives.
## Configure the first variation
On the **Variations** tab, click **New variation**. Enter:
| Field | Value | Explanation |
| ---------------------- | -------------------------------------------------------------------------------------------------- | -------------------------------------------------------------- |
| **Name** | `Default` | Identifies this candidate configuration |
| **Description** | `Answers with a short, direct explanation.` | Records what distinguishes this variation |
| **System prompt** | `Explain the requested Cadenya concept accurately in two or three sentences. Do not use headings.` | Defines the behavior used by every objective on this variation |
| **Model** | Any enabled model | Selects the model that executes the prompt |
| **Temperature** | `0.2` | Keeps answers relatively consistent |
| **Max tool calls** | `0` | Means unlimited, though this variation has no tool assignments |
| **Max sub-objectives** | `0` | Means unlimited, though this variation has no sub-agents |
Click **Create variation**. The variation appears in the left side of the Variations pane. Its details and assignments appear on the right.
## Publish the agent
In the **Details** card, open the **Draft** status menu and select **Publish**.
Publishing makes the agent available to objectives. It does not freeze the configuration. You can set it back to draft before making a controlled change.
## Send an objective
Export your credentials:
```bash theme={null}
export CADENYA_API_KEY='your-api-key'
export CADENYA_WORKSPACE_ID='workspace_your-workspace-id'
```
Use the external ID so application code does not need the generated agent ID.
```typescript TypeScript theme={null}
import Cadenya from '@cadenya/cadenya';
const client = new Cadenya({
apiKey: process.env['CADENYA_API_KEY'],
});
const objective = await client.objectives.create({
workspaceId: process.env['CADENYA_WORKSPACE_ID'],
agentId: 'external_id:concise-helper',
systemPromptData: {},
firstUserMessage: 'What is an objective in Cadenya?',
});
console.log(objective.metadata.id);
```
```bash cURL theme={null}
curl "https://api.cadenya.com/v1/workspaces/${CADENYA_WORKSPACE_ID}/objectives" \
-H "Authorization: Bearer ${CADENYA_API_KEY}" \
-H "Content-Type: application/json" \
-d '{
"agentId": "external_id:concise-helper",
"systemPromptData": {},
"firstUserMessage": "What is an objective in Cadenya?"
}'
```
The create response is a snapshot of the selected agent and variation. Open **Objectives** and select its ID to follow the timeline. Because this agent has no structured output definition, it stops in **Waiting** after its assistant message and can accept another turn.
## What each resource owns
| Resource | Configuration in this guide |
| --------- | ------------------------------------------------------------------------------------ |
| Agent | Name, external ID, description, lifecycle state, and variation selection mode |
| Variation | Prompt, model, temperature, constraints, discovery, compaction, and assignments |
| Objective | One execution, its input data and messages, and a snapshot of the selected variation |
Add another prompt or model and control how Cadenya selects between them.
Attach a tool, a complete tool set, a sub-agent, or memory.
Score completed objectives and influence future variation samples.
Create the same lifecycle from application code.