# Dify on its own vs Dify + LangBot

Source: https://langbot.app/en/compare/dify-vs-dify-with-langbot · Last reviewed: 2026-11-11

Dify builds the AI app; LangBot delivers it to chat platforms. Compare Dify's own publishing options with running a Dify app through LangBot on Discord, Telegram, Slack, WeCom, Lark, DingTalk and QQ.

## In short

Dify is where you build the app: prompts, workflows, tools and knowledge. Its built-in publishing gives you a web app, an embeddable chat widget and a backend API. LangBot is not a replacement for Dify: it calls your published Dify app and handles the messaging platforms, sessions, group-chat rules and delivery.

| | Dify + LangBot | Dify on its own |
| --- | --- | --- |
| Get it running | cloud.langbot.app: connect your Dify app with no deployment, or self-host LangBot | Dify Cloud or self-hosted Dify; chat-app bridges are up to you |
| Set-up help | Workspace Assistant creates the pipeline and bots from a description, after you confirm | — |
| Mix and switch back ends | Run Dify alongside LangBot Models, n8n, Coze or Langflow and switch per pipeline | Dify only |
| Role | Messaging layer: platform adapters, sessions, triggers, delivery | App builder: prompts, Chatflow, Agent, Workflow, knowledge |
| Chat platforms | Discord, Telegram, Slack, LINE, WeCom, Lark, DingTalk, QQ, WeChat Official Account, KOOK, Mattermost and more | Web app, embedded widget and API; messaging channels need extra integration |
| Group chat behaviour | Mention/prefix triggers, access control, rate limits per pipeline | Not a concern of the Dify app itself |
| Same app on many platforms | Bind one pipeline to many bots | Build one integration per channel |
| Licence | Apache 2.0 | Dify Open Source License (Apache 2.0 based, with extra conditions) |

## What LangBot gives you out of the box

### No deployment: cloud.langbot.app

Sign up at [cloud.langbot.app](https://cloud.langbot.app) and build a bot in the browser with a setup wizard. There is a free plan, the public HTTPS URL webhook platforms need is included, and a Dedicated environment is available for production. Self-hosting stays free and unlimited.

### Official model service: LangBot Models

Claude, GPT, Gemini, DeepSeek, Qwen, GLM, Kimi, Grok and more, plus embedding, rerank and speech models, with no model API key to obtain. Cloud plans include monthly credits; self-hosted instances signed in with a LangBot Space account get the models added automatically with a free starter quota. Your own API keys keep working alongside.

### Control your bots in natural language

Agents, a unified API and standard platform events and APIs work together. Every Workspace operation (bots, routing, pipelines, models, knowledge bases, monitoring) is one catalog shared by the HTTP API, the MCP server, the Workspace Assistant and plugins, and adapters expose standard platform APIs such as send, reply, edit and recall messages, query groups and members, and, where the platform allows, mute or remove members. So you can change how a bot behaves by telling the Workspace Assistant, or let an authorized agent act on the platform from a chat instruction. Writes need your confirmation or an explicit permission grant.

### Workspace Assistant

Describe what you want in plain language. The assistant inspects your existing models, knowledge bases and pipelines, proposes a configuration and applies it only after you confirm each change.

### Rich APIs and flexible tools

An HTTP Service API with API keys, LangBot's own MCP server, the `lbctl` CLI and outgoing event webhooks for your systems. Agents can use MCP tools (stdio and HTTP), Skills, a code sandbox (Docker, nsjail or E2B), knowledge bases with rerank, and platform events such as member joins, reactions and card clicks.

### Rich, stable plugin interfaces

Plugins run in a separate Plugin Runtime process and are hot-pluggable. The Python SDK offers Runner, Tool, Command, EventListener, KnowledgeEngine, Parser and Page components, and a Workspace API whose operations a plugin must declare and a member must grant. The [LangBot Space marketplace](https://space.langbot.app/market) lists plugins, MCP servers and Skills, with reviewed Certified Plugins.

### Open source, with official backing

Apache 2.0 with free commercial use, an official managed service, and optional Enterprise Support with contracts and invoices.

## When Dify alone is enough

If your users talk to the assistant in a browser, Dify's web app or embedded widget is the simplest option, and the API covers custom front ends.

## When to add LangBot

- Your users are in chat apps: Slack, Discord, Telegram, WeCom, Lark, DingTalk, QQ or WeChat.
- You want the same Dify app on several platforms without writing an adapter for each.
- You need group-chat rules (respond only when mentioned), per-pipeline access control and conversation monitoring.

LangBot supports Dify Chat Assistant (including Chatflow), Agent and Workflow apps. Setup takes the Dify API base URL and an app API key: see the [Dify runner guide](https://langbot.app/docs/en/usage/pipelines/dify) or a platform recipe such as [Dify on Telegram](https://langbot.app/docs/en/recipes/dify/telegram).

## FAQ

### Does LangBot replace Dify?

No. Dify keeps the prompts, tools, workflows and knowledge. LangBot calls the published Dify app for each message and delivers the answer to the chat platform.

### Can I use self-hosted Dify?

Yes. Use your Dify address with /v1 as the base URL. If both run in Docker on one host, put them on a shared Docker network.

## Further reading

- [Dify](https://dify.ai)
- [Dify runner guide](https://langbot.app/docs/en/usage/pipelines/dify)