Prompts

Prompts are versioned templates stored in the workspace. You edit them in the dashboard or API, then fetch the right version at runtime by handle — latest, production, or a specific version number. Actx0 does not execute the prompt. Your agent loads the template, substitutes variables, and sends it to your model along with memories and knowledge hits.

The mental model

Without prompt management	With Actx0 prompts
Prompt text hardcoded in deploys	Change copy without shipping a new binary
No history when a change regresses	Every save is a numbered version with a commit message
Staging and prod share one string	Promote a version to `production` when it is ready

Prompt (name + handle)
  ├── v1  labels: []
  ├── v2  labels: [latest]
  └── v3  labels: [latest]  production: true

A prompt is the parent (name, handle, description). A version holds type, content, optional config, labels, and the production flag.

Handles and versions

The handle is derived from the name: lowercase, dashes for spaces, unique in the workspace. "Support system prompt" becomes support-system-prompt. That handle is what runtime code should call.

Selector	What you get
`latest` (default)	The newest version. Creating a version moves `latest` onto it
`production`	The version you promoted for live traffic
`1` or `v1`	A specific version number

curl -X GET \
  "https://app.actx0.com/api/v1/workspaces/{workspaceId}/promptsByName/support-system-prompt?version=production" \
  -H "X-Access-Key: $ACTX0_ACCESS_KEY" \
  -H "Accept: application/json"

Fetch production in live agents and latest in staging. Promote a version in the dashboard or by updating it with "production": true. Only one version is production at a time.

Types, variables, and config

Field	Purpose
`type`	`text` (single string) or `chat` (multi-turn template)
`content`	The template body
`config`	Optional JSON string for model params, tools, or app-specific settings
`commitMessage`	Human note stored with the version
`production`	Set on create or update to mark this version live

Insert variables as {{variable}}. Names must be lowercase with underscores, for example {{user_name}} or {{ticket_id}}. Actx0 stores the placeholders; your app substitutes values before calling the model. Treat versions as immutable in the product workflow: saving in the dashboard creates a new version rather than silently rewriting history. You can still archive or delete a version, or delete the whole prompt and all of its versions.

Create a prompt

Creating a prompt also creates version 1.

curl -X POST \
  "https://app.actx0.com/api/v1/workspaces/{workspaceId}/prompts" \
  -H "X-Access-Key: $ACTX0_ACCESS_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json" \
  -d '{
    "name": "Support system prompt",
    "description": "Primary system prompt for support agents",
    "type": "text",
    "content": "You are a helpful support agent. Address the user as {{user_name}}.",
    "commitMessage": "Initial version",
    "production": true
  }'

Later, POST a new version on the same prompt id when the copy changes. Prompt count is limited by your workspace plan.

Build against this flow

Call prompts by handle, not by UUID, so deploys stay stable when you add versions.
Keep production on a known-good version; experiment on latest.
Pair the fetched template with memory search and knowledge search in the same request path.
Put model hyperparameters in config if your runtime reads them; Actx0 does not apply config itself.
Do not put secrets in prompt content.

Next steps

Create Prompt

First version, handle, and production flag

Get Prompt by Name

Resolve latest, production, or a version number

Agents

Where prompts are used at session time

Knowledge

Ground the same prompt with retrieved documents

​Prompts

​The mental model

​Handles and versions

​Types, variables, and config

​Create a prompt

​Build against this flow

​Next steps

Create Prompt

Get Prompt by Name

Agents

Knowledge

Prompts

The mental model

Handles and versions

Types, variables, and config

Create a prompt

Build against this flow

Next steps