Tools

Point almyty at an API and every operation becomes a typed tool your agents can call: validated inputs, a description the model reads to decide when to use it, and a source it traces back to. Import a schema and you get a tool per operation. Need custom logic instead? Write one by hand from HTTP, GraphQL, JavaScript, an LLM prompt, or an npm SDK.

Tools list showing typed tools with name, type, status, and source API

Two ways to get tools

Import an API. On the APIs page, import a schema and almyty creates a tool for each operation, with parameters, types, and validation already filled in:

Schema	What becomes a tool
OpenAPI	Each path plus method
GraphQL	Each query and mutation
SOAP	Each WSDL operation
Protobuf	Each RPC method

Create one by hand. Click Create Tool and pick an execution method. Give it a name and description (the description is what agents see when deciding whether to call it), define input parameters with the JSON Schema builder, then configure the method.

Create tool dialog showing the five execution methods

Tool types

Type	Use it to	Docs
HTTP	Call any REST endpoint with method, URL, headers, and body	HTTP tools
GraphQL	Run a query or mutation against a GraphQL endpoint	GraphQL tools
JavaScript	Run sandboxed JS/TS with npm packages for logic and transforms	JavaScript tools
LLM	Generate output from a prompt template using a configured model	LLM tools
SDK	Import an npm package and generate tools from its exported functions	SDK tools
MCP	Consume any remote MCP server; its tools become native almyty tools	MCP tools

Open a tool to configure it and run it against real inputs in the tool tester before an agent ever touches it.

Tool detail with configuration and the tool tester

Put tools in front of an LLM

Assign tools to a gateway, then expose that gateway over MCP from your terminal. Any MCP-compatible client (Claude, an IDE, your own agent) sees them as native tools:


$ npx @almyty/auth login
$ npx @almyty/mcp-server acme/petstore

See the MCP server CLI for gateway setup and client config.

Tool lifecycle

Status	Meaning
`active`	Available for execution and gateway assignment
`inactive`	Exists but cannot be executed
`deprecated`	Marked for removal, still functional
`error`	Has a configuration issue that needs attention

Export formats

Any tool exports to:

SKILL.md for Agent Skills injection
CLI as a Bash or Node.js script
SDK as generated code for programmatic use

See the individual tool type pages for method-specific configuration.