@marktoflow/integrations

39 native service integrations and AI agent adapters with tool calling, structured output, and local LLM support.

Part of marktoflow — open-source AI workflow automation.

Quick Start

npm install @marktoflow/integrations

Use in workflows:

tools:
  slack:
    sdk: '@slack/web-api'
    auth:
      token: '${SLACK_BOT_TOKEN}'

steps:
  - action: slack.chat.postMessage
    inputs:
      channel: '#general'
      text: 'Hello from marktoflow!'

Or programmatically:

import { SlackInitializer } from '@marktoflow/integrations';
import { SDKRegistry } from '@marktoflow/core';

const registry = new SDKRegistry();
await registry.registerSDK(SlackInitializer);
const slack = await registry.loadSDK('slack', { auth: { token: process.env.SLACK_BOT_TOKEN } });
await registry.executeAction('slack', 'chat.postMessage', slack, { channel: '#general', text: 'Hello!' });

Features

• 39 native SDK integrations with full TypeScript types
• 7 AI agent adapters — OpenAI, Claude, Copilot, Codex, OpenCode, Ollama + any OpenAI-compatible endpoint
• Tool calling / function calling — Agentic loops where AI models invoke tools autonomously
• Structured output — JSON mode and JSON Schema validation for reliable AI responses
• Local LLM support — llama.cpp, VLLM, LM Studio, LocalAI with auto model detection
• Input validation via Zod schemas for every action
• Automatic retry with circuit breakers and exponential backoff
• Credential encryption (AES-256-GCM) and OAuth token refresh
• 256+ contract tests across 28 services (MSW-based, no API keys needed)

AI Agent Adapters

| Agent | Setup | Tool Calling | Structured Output | |-------|-------|:------------:|:-----------------:| | OpenAI | OPENAI_API_KEY | ✅ | ✅ | | Local LLM (llama.cpp, VLLM, LM Studio) | base_url + api_key: dummy | ✅ | ✅ | | GitHub Copilot | copilot auth login | ✅ | — | | Claude Agent | Claude CLI or ANTHROPIC_API_KEY | ✅ | — | | OpenAI Codex | Codex CLI | ✅ | — | | OpenCode | opencode /connect | ✅ | — | | Ollama | Local install | ✅ | — |

OpenAI / Local LLM Methods

| Method | Description | |--------|-------------| | ai.generate | Simple text generation | | ai.chatCompletion | Full chat completion with tools and response_format | | ai.chatWithTools | Agentic tool-calling loop (model calls tools → execute → repeat) | | ai.generateJSON | Chat with JSON mode — returns parsed JSON | | ai.generateStructured | Chat with JSON Schema validation | | ai.chatStream | Streaming chat completion | | ai.embeddings | Generate text embeddings | | ai.listModels | List available models | | ai.autoDetectModel | Auto-detect model from server |

Tool Calling Example

tools:
  ai:
    sdk: openai
    auth:
      base_url: http://localhost:8000/v1  # llama.cpp, VLLM, etc.
      api_key: dummy
    options:
      model: auto

steps:
  - action: ai.chatWithTools
    inputs:
      messages:
        - role: user
          content: "What's the weather in London?"
      tools:
        - type: function
          function:
            name: get_weather
            description: Get current weather
            parameters:
              type: object
              properties:
                city: { type: string }
              required: [city]
      maxTurns: 5

Service Reference

| Service | Category | Key Actions | |---------|----------|-------------| | Slack | Communication | chat.postMessage, conversations.list, users.list | | Teams | Communication | sendMessage, createChannel, createMeeting | | Discord | Communication | sendMessage, editMessage, deleteMessage | | Telegram | Communication | sendMessage, sendPhoto, sendDocument | | WhatsApp | Communication | sendMessage, sendTemplate, sendMedia | | Twilio | Communication | sendSMS, makeCall, sendWhatsApp | | Gmail | Email | users.messages.send, users.messages.list | | Outlook | Email | sendMail, listMessages, listCalendarEvents | | SendGrid | Email | sendEmail, sendMultiple | | Mailchimp | Email | addMember, createCampaign, sendCampaign | | Google Sheets | Productivity | getValues, updateValues, appendValues | | Google Calendar | Productivity | listEvents, createEvent, deleteEvent | | Google Drive | Productivity | listFiles, uploadFile, downloadFile | | Google Docs | Productivity | getDocument, createDocument, appendText | | Notion | Knowledge | databases.query, pages.create, blocks.children.append | | Confluence | Knowledge | getPage, createPage, updatePage | | Jira | Project Mgmt | issues.createIssue, issues.searchIssues | | Linear | Project Mgmt | createIssue, updateIssue, listIssues | | Asana | Project Mgmt | createTask, updateTask, getTasksInProject | | Trello | Project Mgmt | createCard, updateCard, addChecklistToCard | | GitHub | Developer | pulls.create, issues.create, repos.get | | Airtable | Developer | select, create, update, delete | | Stripe | Payments | createCustomer, createPaymentIntent, createSubscription | | Shopify | Commerce | getProducts, createOrder, updateInventoryLevel | | Zendesk | Support | createTicket, updateTicket, search | | Dropbox | Storage | uploadFile, downloadFile, listFolder | | AWS S3 | Storage | uploadObject, getObject, listObjects | | Supabase | Database | select, insert, update, rpc | | PostgreSQL | Database | query, insert, update, delete | | MySQL | Database | query, insert, update, delete | | HTTP | Universal | request (any REST API) |

Creating Custom Integrations

import type { SDKInitializer } from '@marktoflow/core';

export const MyServiceInitializer: SDKInitializer = {
  name: 'myservice',
  async initialize(config) {
    return new MyServiceClient(config.auth.apiKey);
  },
  actions: {
    doSomething: async (sdk, inputs) => sdk.doSomething(inputs),
  },
};

For per-service setup details (environment variables, OAuth, examples), see the full documentation.

Contributing

See the contributing guide.

License

AGPL-3.0