
Research
/Security News
Mini Shai-Hulud Campaign Hits Red Hat Cloud Services npm Packages
A mini Shai-Hulud campaign compromised Red Hat Cloud Services npm packages to steal developer and CI/CD secrets during installation.
@loopstack/custom-tool-example-module
Advanced tools
A complete example demonstrating how to implement and use a custom tool in a workflow
A module for the Loopstack AI automation framework.
This module provides a complete example demonstrating how to implement and use custom tools in a Loopstack workflow.
Custom tools are the building blocks of Loopstack automations. This module serves as a hands-on reference for developers learning how to extend Loopstack with their own functionality.
By exploring this example, you'll understand:
BaseTool with a call() method@Tool decorator@InjectTool()wait: true on transitions for manual triggersThis is a great starting point before building your own custom tools.
See SETUP.md for installation and setup instructions.
A simple tool that maintains internal state across calls. It extends BaseTool and implements a call() method:
import { BaseTool, Tool, ToolResult } from '@loopstack/common';
@Tool({
uiConfig: {
description: 'Counter tool.',
},
})
export class CounterTool extends BaseTool {
count: number = 0;
call(_args?: object): Promise<ToolResult<number>> {
this.count++;
return Promise.resolve({ data: this.count });
}
}
The count property persists across calls within the same workflow execution, so each call increments the counter.
A tool that accepts typed arguments via a Zod schema and uses NestJS dependency injection for services:
import { Inject } from '@nestjs/common';
import { z } from 'zod';
import { BaseTool, Tool, ToolResult } from '@loopstack/common';
import { MathService } from '../services/math.service';
const MathSumSchema = z
.object({
a: z.number(),
b: z.number(),
})
.strict();
type MathSumArgs = z.infer<typeof MathSumSchema>;
@Tool({
uiConfig: {
description: 'Math tool calculating the sum of two arguments by using an injected service.',
},
schema: MathSumSchema,
})
export class MathSumTool extends BaseTool {
@Inject()
private mathService: MathService;
call(args: MathSumArgs): Promise<ToolResult<number>> {
const sum = this.mathService.sum(args.a, args.b);
return Promise.resolve({ data: sum });
}
}
The schema option on @Tool validates incoming arguments. The injected MathService is a standard NestJS injectable:
@Injectable()
export class MathService {
public sum(a: number, b: number) {
return a + b;
}
}
The workflow extends BaseWorkflow<TArgs> with a typed argument object. The schema is defined in the @Workflow decorator:
@Workflow({
uiConfig: __dirname + '/custom-tool-example.ui.yaml',
schema: z
.object({
a: z.number().default(1),
b: z.number().default(2),
})
.strict(),
})
export class CustomToolExampleWorkflow extends BaseWorkflow<{ a: number; b: number }> {
@InjectTool() private counterTool: CounterTool;
@InjectTool() private mathTool: MathSumTool;
total?: number;
}
Call tools via this.tool.call(args) inside transition methods. Store the result as an instance property:
@Initial({ to: 'waiting_for_user' })
async calculate(args: { a: number; b: number }) {
const calcResult = await this.mathTool.call({ a: args.a, b: args.b });
this.total = calcResult.data as number;
await this.repository.save(MessageDocument, {
role: 'assistant',
content: `Tool calculation result:\n${args.a} + ${args.b} = ${this.total}`,
});
await this.repository.save(MessageDocument, {
role: 'assistant',
content: `Alternatively, using workflow method:\n${args.a} + ${args.b} = ${this.sum(args.a, args.b)}`,
});
}
The counter tool increments on each call, demonstrating that tool state persists within a workflow execution:
const c1 = await this.counterTool.call();
const c2 = await this.counterTool.call();
const c3 = await this.counterTool.call();
await this.repository.save(MessageDocument, {
role: 'assistant',
content: `Counter before pause: ${c1.data}, ${c2.data}, ${c3.data}\n\nPress Next to continue...`,
});
Use wait: true on a transition to pause the workflow until it is manually triggered (e.g., by user input):
@Transition({ from: 'waiting_for_user', to: 'resumed', wait: true })
async userContinue() {}
The workflow pauses at the waiting_for_user state until an external signal triggers the userContinue transition.
A @Final method can return data as the workflow output:
@Final({ from: 'resumed' })
async continueCount(): Promise<{ total: number | undefined }> {
const c4 = await this.counterTool.call();
const c5 = await this.counterTool.call();
const c6 = await this.counterTool.call();
await this.repository.save(MessageDocument, {
role: 'assistant',
content: `Counter after resume: ${c4.data}, ${c5.data}, ${c6.data}\n\nIf state persisted, this should be 4, 5, 6.`,
});
return { total: this.total };
}
After resuming, the counter continues from where it left off (4, 5, 6), demonstrating that tool state survives a wait pause.
Define private methods for reusable logic within the workflow:
private sum(a: number, b: number) {
return a + b;
}
import { z } from 'zod';
import { BaseWorkflow, Final, Initial, InjectTool, Transition, Workflow } from '@loopstack/common';
import { MessageDocument } from '@loopstack/common';
import { MathSumTool } from '../tools';
import { CounterTool } from '../tools';
@Workflow({
uiConfig: __dirname + '/custom-tool-example.ui.yaml',
schema: z
.object({
a: z.number().default(1),
b: z.number().default(2),
})
.strict(),
})
export class CustomToolExampleWorkflow extends BaseWorkflow<{ a: number; b: number }> {
@InjectTool() private counterTool: CounterTool;
@InjectTool() private mathTool: MathSumTool;
total?: number;
@Initial({ to: 'waiting_for_user' })
async calculate(args: { a: number; b: number }) {
const calcResult = await this.mathTool.call({ a: args.a, b: args.b });
this.total = calcResult.data as number;
await this.repository.save(MessageDocument, {
role: 'assistant',
content: `Tool calculation result:\n${args.a} + ${args.b} = ${this.total}`,
});
await this.repository.save(MessageDocument, {
role: 'assistant',
content: `Alternatively, using workflow method:\n${args.a} + ${args.b} = ${this.sum(args.a, args.b)}`,
});
const c1 = await this.counterTool.call();
const c2 = await this.counterTool.call();
const c3 = await this.counterTool.call();
await this.repository.save(MessageDocument, {
role: 'assistant',
content: `Counter before pause: ${c1.data}, ${c2.data}, ${c3.data}\n\nPress Next to continue...`,
});
}
@Transition({ from: 'waiting_for_user', to: 'resumed', wait: true })
async userContinue() {}
@Final({ from: 'resumed' })
async continueCount(): Promise<{ total: number | undefined }> {
const c4 = await this.counterTool.call();
const c5 = await this.counterTool.call();
const c6 = await this.counterTool.call();
await this.repository.save(MessageDocument, {
role: 'assistant',
content: `Counter after resume: ${c4.data}, ${c5.data}, ${c6.data}\n\nIf state persisted, this should be 4, 5, 6.`,
});
return { total: this.total };
}
private sum(a: number, b: number) {
return a + b;
}
}
This workflow uses the following Loopstack modules:
@loopstack/common - Base classes, decorators, and tool injection@loopstack/common - Provides MessageDocument for chat messagesAuthor: Jakob Klippel
License: MIT
FAQs
A complete example demonstrating how to implement and use a custom tool in a workflow
The npm package @loopstack/custom-tool-example-module receives a total of 44 weekly downloads. As such, @loopstack/custom-tool-example-module popularity was classified as not popular.
We found that @loopstack/custom-tool-example-module demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?

Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.

Research
/Security News
A mini Shai-Hulud campaign compromised Red Hat Cloud Services npm packages to steal developer and CI/CD secrets during installation.

Research
/Security News
The North Korean malware loader hides in a Packagist-listed package and its GitHub branch to fetch and execute remote code in a likely Contagious Interview-style lure.

Security News
The Rust project is moving toward formal rules on LLM use in contributions after months of internal debate over maintainer burden, code quality, and contributor experience.