What are Tools?
Tools extend Agent and Flow capabilities by providing functions for specific tasks. Runtype supports built-in tools, custom JavaScript tools, external HTTP tools, Flows as tools, runtime tools, and MCP servers.
Tool types
Built-in tools
Ready-to-use integrations provided by Runtype:
DALL-E: Generate images from text descriptions
Exa: Search the web for current information
Firecrawl: Scrape and extract structured data from websites
OpenAI Web Search: Search the web using OpenAI's search capabilities
Anthropic Web Search: Search the web using Anthropic's search capabilities
Anthropic Web Fetch: Fetch web content using Anthropic's fetch tool
Semantic Search: Search vector knowledge bases
Built-in tools appear when you add tools to a prompt or Agent and choose a compatible model. Configure API keys as needed: Firecrawl and Exa under Settings → Integrations; for DALL-E and other provider tools, set up the provider (e.g. OpenAI) under Settings → Models.
Custom tools
JavaScript functions you write for specialized tasks. Execute custom logic within Runtype's sandboxed environment.
Use for: Data transformations, calculations, validation, business logic
External tools
HTTP API endpoints your Agents and Flows can call. Integrate with external services and databases.
Use for: Third-party APIs, microservices, databases, legacy systems
Flow tools
Turn any Flow into a tool that Agents and Flows can invoke. Reuse complex workflows as building blocks.
Use for: Multi-step operations, orchestrated tasks, reusable workflows
Runtime tools
Tools defined inline with API requests (e.g. when calling the dispatch API) without saving them first. They support the same types as saved tools: external HTTP, custom code, and Flow.
Use for: Dynamic or temporary tool config, SDK-driven Flows, testing, or when you don't want to store tool definitions in the platform
MCP servers
Connect MCP (Model Context Protocol) servers to expose additional tools to your Agents and Flows. Runtype supports both built-in and custom MCP servers.
Use for: Extending tool capabilities with standardized MCP integrations
How Agents use tools
When you add tools to an Agent:
Agent reads tool descriptions and parameters
Agent decides which tool to use based on the task
Agent generates appropriate parameters
Tool executes and returns results
Agent uses results to continue reasoning
See Agent tools for details on tool selection.
Tool structure
Every tool has:
Name: Unique identifier
Description: What the tool does and when to use it
Parameters: Input schema with types and descriptions
Implementation: Code, URL, or Flow reference
Return type: What the tool outputs
Managing tools
Tools are created and managed in the Tools section:
Click Tools in the sidebar
View all available tools (saved custom, external, and Flow tools)
Click a tool to edit or view details
Next steps
Built-in tools for using Exa, Firecrawl, and DALL-E
Creating custom tools to build JavaScript-based tools
Creating external tools for HTTP API integration
Runtime tools for advanced execution environments
Agent tools to understand how Agents select and use tools