Glyph Types
Deep dive into the three types of Runnable Digital Assets.
Overview
| Type | Description | Use Case | Pricing |
|---|---|---|---|
| Prompts | LLM-based generation | Text analysis, content creation | Model tiers |
| Agents | Workflow orchestration | Automation, multi-step tasks | Custom |
| Tools | API proxies | Data access, integrations | Custom |
Prompts
What They Are
Prompts use large language models to generate text based on:
- A prompt template
- User-provided inputs
- Selected model tier
How They Work
User Input → Prompt Template → LLM → Output- User fills input form
- Inputs are injected into template
- Template sent to LLM
- Response returned to user
Key Features
- Model selection - Choose AI model tier
- Template variables - Dynamic input injection
- Examples - Pre-generated output samples
- Source purchase - Buy the prompt template
Best For
- Text generation and analysis
- Creative content
- Technical documentation
- Planning and strategy
Agents
What They Are
Agents orchestrate workflows by communicating with external services via webhooks.
How They Work
User Input → Webhook → External Service → Result- User provides inputs
- Agent calls webhook endpoint
- External service processes (n8n, CrewAI, etc.)
- Result returned to user
Key Features
- Webhook integration - Connect to any HTTP endpoint
- Async support - Long-running task polling
- Multi-step - Chain complex workflows
- Retries - Automatic failure handling
Best For
- Automation workflows
- Multi-step processes
- External service orchestration
- Scheduled tasks
Tools
What They Are
Tools wrap external APIs with payment handling and user-friendly inputs.
How They Work
User Input → Transform → API Call → Transform → Output- User provides inputs
- Request is transformed
- External API is called
- Response is transformed
- Result returned to user
Key Features
- API wrapping - Any HTTP API
- Authentication - Secure credential handling
- Transformation - Request/response formatting
- Rate limiting - Usage control
Best For
- Data access
- External service integration
- API aggregation
- Service proxying
Comparison
Execution Model
| Aspect | Prompts | Agents | Tools |
|---|---|---|---|
| Engine | LLM API | Webhook | HTTP API |
| Latency | 2-60s | Variable | 1-10s |
| Output | Text | Any | JSON/Text |
| Async | No | Yes | No |
Creator Requirements
| Aspect | Prompts | Agents | Tools |
|---|---|---|---|
| Needs | Prompt engineering | Webhook endpoint | API access |
| Complexity | Low | Medium | Medium |
| Hosting | None | External | None |
| Secrets | None | Optional | API keys |
Pricing
| Aspect | Prompts | Agents | Tools |
|---|---|---|---|
| Model | Tier-based | Custom | Custom |
| Range | $0.02-0.10 | $0.05+ | $0.01+ |
| Source | Optional sale | N/A | N/A |
Choosing a Type
Create a Prompt When:
- Your capability is text-based
- Output comes from AI generation
- You want tiered pricing
- No external services needed
Create an Agent When:
- You need multi-step workflows
- Tasks are long-running
- External services are involved
- Complex orchestration needed
Create a Tool When:
- Wrapping an existing API
- Data access is the goal
- Simple request/response
- Authentication needed
Technical Details
Prompt Internals
interface PromptConfig {
promptTemplate: string
systemPrompt?: string
allowedModels: ModelOption[]
defaultModelId: string
sourcePrice?: number
examples?: Example[]
}Agent Internals
interface AgentConfig {
webhookUrl: string
webhookMethod: 'GET' | 'POST'
webhookHeaders?: Record<string, string>
timeoutSeconds: number
maxRetries: number
pollUrl?: string
pollIntervalMs?: number
}Tool Internals
interface ToolConfig {
apiEndpoint: string
apiMethod: 'GET' | 'POST' | 'PUT' | 'DELETE'
apiHeaders?: Record<string, string>
authType: 'none' | 'api_key' | 'bearer' | 'basic'
authConfig?: Record<string, string>
requestTransform?: string
responseTransform?: string
rateLimitPerMinute?: number
}Related Topics
Last updated on: