Skip to main content

Agentspec

Agent Runtimes

Agent Runtimes / types/agentspecs / Agentspec

Interface: Agentspec

Defined in: src/types/agentspecs.ts:27

Specification for an AI agent.

Defines the configuration for a reusable agent template that can be instantiated as an Agent Runtime.

Properties​

advanced?​

optional advanced?: AgentAdvancedConfig

Defined in: src/types/agentspecs.ts:99

Advanced settings (cost_limit, time_limit, max_iterations, validation)

authorizationPolicy?​

optional authorizationPolicy?: string

Defined in: src/types/agentspecs.ts:101

Authorization policy

codemode?​

optional codemode?: AgentCodemodeConfig

Defined in: src/types/agentspecs.ts:95

Codemode configuration (enabled, token_reduction, speedup)

color?​

optional color?: string

Defined in: src/types/agentspecs.ts:67

Theme color for the agent (hex code)

description​

description: string

Defined in: src/types/agentspecs.ts:35

Agent description

disableToolApprovals?​

optional disableToolApprovals?: boolean

Defined in: src/types/agentspecs.ts:55

Disable tool approvals for this spec (default: false).

emoji?​

optional emoji?: string

Defined in: src/types/agentspecs.ts:65

Emoji identifier for the agent

enabled​

enabled: boolean

Defined in: src/types/agentspecs.ts:43

Whether the agent is enabled

environmentName​

environmentName: string

Defined in: src/types/agentspecs.ts:61

Runtime environment name for this agent

evals?​

optional evals?: AgentEvalConfig[]

Defined in: src/types/agentspecs.ts:93

Evaluation configurations

frontendRenderTools?​

optional frontendRenderTools?: FrontendRenderToolSpec[]

Defined in: src/types/agentspecs.ts:59

Bindings of backend tools to frontend renderers (tool name + css)

frontendTools?​

optional frontendTools?: FrontendToolSpec[]

Defined in: src/types/agentspecs.ts:57

Frontend tool sets available to this agent

goal?​

optional goal?: string

Defined in: src/types/agentspecs.ts:79

User-facing objective for the agent

guardrails?​

optional guardrails?: GuardrailSpec[]

Defined in: src/types/agentspecs.ts:91

Guardrail configurations

icon?​

optional icon?: string

Defined in: src/types/agentspecs.ts:63

Icon identifier or URL for the agent

id​

id: string

Defined in: src/types/agentspecs.ts:29

Unique agent identifier

inferenceProvider?​

optional inferenceProvider?: "datalayer" | "local"

Defined in: src/types/agentspecs.ts:47

Inference provider routing strategy

mcpServers​

mcpServers: MCPServer[]

Defined in: src/types/agentspecs.ts:49

MCP servers used by this agent

mcpServerTools?​

optional mcpServerTools?: AgentMCPServerToolConfig[]

Defined in: src/types/agentspecs.ts:89

MCP server tool configurations with approval settings

memory?​

optional memory?: string

Defined in: src/types/agentspecs.ts:105

Memory backend identifier (e.g., 'ephemeral', 'mem0', 'memu', 'simplemem')

model?​

optional model?: string

Defined in: src/types/agentspecs.ts:45

AI model identifier to use for this agent

modelConfig?​

optional modelConfig?: AgentModelConfig

Defined in: src/types/agentspecs.ts:87

Model configuration (temperature, max_tokens)

name​

name: string

Defined in: src/types/agentspecs.ts:33

Display name for the agent

notifications?​

optional notifications?: AgentNotificationConfig

Defined in: src/types/agentspecs.ts:103

Notification configuration (email, slack)

output?​

optional output?: AgentOutputConfig

Defined in: src/types/agentspecs.ts:97

Output configuration (type/formats, template)

parameters?​

optional parameters?: Record<string, any>

Defined in: src/types/agentspecs.ts:118

JSON schema for launch-time parameter values.

postHooks?​

optional postHooks?: object

Defined in: src/types/agentspecs.ts:112

Post-stop hooks (sandbox cleanup code).

sandbox?​

optional sandbox?: string | string[]

preHooks?​

optional preHooks?: object

Defined in: src/types/agentspecs.ts:107

Pre-launch hooks (package installs and sandbox code).

packages?​

optional packages?: string[]

sandbox?​

optional sandbox?: string | string[]

protocol?​

optional protocol?: string

Defined in: src/types/agentspecs.ts:81

Communication protocol (e.g., 'ag-ui', 'acp', 'a2a', 'vercel-ai')

sandboxVariant?​

optional sandboxVariant?: string

Defined in: src/types/agentspecs.ts:77

Sandbox variant to use for this agent (e.g. 'eval', 'jupyter', 'kaggle').

skills​

skills: SkillSpec[]

Defined in: src/types/agentspecs.ts:51

Skills available to this agent

subagents?​

optional subagents?: SubAgentsConfig

Defined in: src/types/agentspecs.ts:120

Subagent delegation configuration.

suggestions?​

optional suggestions?: string[]

Defined in: src/types/agentspecs.ts:69

Chat suggestions to show users what this agent can do

systemPrompt?​

optional systemPrompt?: string

Defined in: src/types/agentspecs.ts:37

System prompt for the agent

systemPromptCodemodeAddons?​

optional systemPromptCodemodeAddons?: string

Defined in: src/types/agentspecs.ts:39

System prompt addons when codemode is enabled

tags​

tags: string[]

Defined in: src/types/agentspecs.ts:41

Tags for categorization

toolHooks?​

optional toolHooks?: Record<string, any>

Defined in: src/types/agentspecs.ts:116

Per-tool-call hooks (authorization/audit integration).

tools?​

optional tools?: ToolSpec[]

Defined in: src/types/agentspecs.ts:53

Runtime tools available to this agent

trigger?​

optional trigger?: AgentTriggerConfig

Defined in: src/types/agentspecs.ts:85

Trigger configuration (type, cron, event source, prompt)

uiExtension?​

optional uiExtension?: string

Defined in: src/types/agentspecs.ts:83

UI extension type (e.g., 'a2ui', 'mcp-apps')

version?​

optional version?: string

Defined in: src/types/agentspecs.ts:31

Version

welcomeDocument?​

optional welcomeDocument?: string

Defined in: src/types/agentspecs.ts:75

Path to Lexical document to show on agent creation

welcomeMessage?​

optional welcomeMessage?: string

Defined in: src/types/agentspecs.ts:71

Welcome message shown when agent starts

welcomeNotebook?​

optional welcomeNotebook?: string

Defined in: src/types/agentspecs.ts:73

Path to Jupyter notebook to show on agent creation