Advancedaiautomationmessaging

AI Agent with ToolsAI Workflow

An autonomous AI agent that can search the web, query a database, and send Slack messages — all through LLM function calling. Define tools once, let the model decide when to use them.

Prerequisites

Environment variables

OPENAI_KEYSLACK_BOT_TOKEN

The agent loops up to maxIterations times. Monitor token usage in production.

Install

$npx radzor@latest recipe add ai-agent

AI Prompt

“Run `npx radzor@latest add function-calling web-scraper embeddings-store slack-bot` to install 4 Radzor components. Then read components/radzor/function-calling/radzor.manifest.json, components/radzor/web-scraper/radzor.manifest.json, components/radzor/embeddings-store/radzor.manifest.json, components/radzor/slack-bot/radzor.manifest.json and each component's llm/integration.md. Wire them together to an autonomous AI agent that can search the web, query a database, and send Slack messages — all through LLM function calling. Define tools once, let the model decide when to use them. Use the manifest's inputs (check envVar for required environment variables), outputs (check fields for object shapes), composability (check mapField for field extraction), and actions — don't invent custom interfaces.”

Paste this into Claude Code, Cursor, Windsurf, or any AI coding agent.

Pipeline

FunctionCalling

LLM decides which tools to call

→

↓

tool calls

WebScraper

Fetches web pages on demand

→

↓

page content

EmbeddingsStore

Searches internal knowledge base

→

↓

search results

SlackBot

Posts messages to Slack

Scaffolded Code

ai-agent-recipe.ts

// npx radzor@latest add function-calling web-scraper embeddings-store slack-bot
import { FunctionCalling } from "./components/radzor/function-calling"
import { WebScraper }      from "./components/radzor/web-scraper"
import { EmbeddingsStore } from "./components/radzor/embeddings-store"
import { SlackBot }        from "./components/radzor/slack-bot"

const agent   = new FunctionCalling({ provider: "openai", apiKey: process.env.OPENAI_KEY!, model: "gpt-4o", maxIterations: 5, systemPrompt: "You are a helpful assistant with access to tools." })
const scraper = new WebScraper({ timeout: 10000 })
const kb      = new EmbeddingsStore({ provider: "openai", apiKey: process.env.OPENAI_KEY! })
const slack   = new SlackBot({ botToken: process.env.SLACK_BOT_TOKEN! })

// Define tools the agent can use
agent.define([
  {
    name: "search_web",
    description: "Fetch and return the text content of a web page",
    schema: { url: { type: "string", description: "URL to fetch" } },
    handler: async (input: { url: string }) => scraper.fetchHtml(input.url),
  },
  {
    name: "search_knowledge_base",
    description: "Search the internal knowledge base by semantic similarity",
    schema: { query: { type: "string" }, topK: { type: "number" } },
    handler: async (input: { query: string; topK: number }) => kb.search(input.query, input.topK),
  },
  {
    name: "send_slack_message",
    description: "Send a message to a Slack channel",
    schema: { channel: { type: "string" }, text: { type: "string" } },
    handler: async (input: { channel: string; text: string }) => slack.sendMessage(input.channel, input.text),
  },
])

// Run the agent
const { response, toolCallLog } = await agent.run("Find our refund policy and post a summary in #support")
console.log(`Agent used ${toolCallLog.length} tools: ${toolCallLog.map(t => t.tool).join(", ")}`)

Components used

FunctionCallingLLM decides which tools to call

View

WebScraperFetches web pages on demand

View

EmbeddingsStoreSearches internal knowledge base

View

SlackBotPosts messages to Slack

View

LLM tip

Pass all 4 radzor.manifest.json files to your agent at once. It will read the outputs of each step and match them against the inputs of the next — wiring the full pipeline without any extra instructions.

function-calling/manifest.json web-scraper/manifest.json embeddings-store/manifest.json slack-bot/manifest.json

// npx radzor@latest add function-calling web-scraper embeddings-store slack-bot import { FunctionCalling } from "./components/radzor/function-calling" import { WebScraper } from "./components/radzor/web-scraper" import { EmbeddingsStore } from "./components/radzor/embeddings-store" import { SlackBot } from "./components/radzor/slack-bot" const agent = new FunctionCalling({ provider: "openai", apiKey: process.env.OPENAI_KEY!, model: "gpt-4o", maxIterations: 5, systemPrompt: "You are a helpful assistant with access to tools." }) const scraper = new WebScraper({ timeout: 10000 }) const kb = new EmbeddingsStore({ provider: "openai", apiKey: process.env.OPENAI_KEY! }) const slack = new SlackBot({ botToken: process.env.SLACK_BOT_TOKEN! }) // Define tools the agent can use agent.define([ { name: "search_web", description: "Fetch and return the text content of a web page", schema: { url: { type: "string", description: "URL to fetch" } }, handler: async (input: { url: string }) => scraper.fetchHtml(input.url), }, { name: "search_knowledge_base", description: "Search the internal knowledge base by semantic similarity", schema: { query: { type: "string" }, topK: { type: "number" } }, handler: async (input: { query: string; topK: number }) => kb.search(input.query, input.topK), }, { name: "send_slack_message", description: "Send a message to a Slack channel", schema: { channel: { type: "string" }, text: { type: "string" } }, handler: async (input: { channel: string; text: string }) => slack.sendMessage(input.channel, input.text), }, ]) // Run the agent const { response, toolCallLog } = await agent.run("Find our refund policy and post a summary in #support") console.log(`Agent used ${toolCallLog.length} tools: ${toolCallLog.map(t => t.tool).join(", ")}`)