← Compare LangBot

Coze publishing vs Coze + LangBot

Coze can publish agents to its own list of channels. Compare that with calling a Coze agent through LangBot to reach Discord, Telegram, Slack, WeCom, Lark, DingTalk, QQ and more from one place.

Last reviewed:

In short

Coze publishes agents to a set of channels that Coze chooses, and the list differs between coze.com and coze.cn. LangBot calls a Coze agent through the Coze API, so the same agent can answer on any platform LangBot supports, with LangBot's group-chat rules, access control and monitoring.

Coze + LangBotCoze publishing
Get it runningcloud.langbot.app: connect your Coze agent with no deploymentPublish from Coze
Mix and switch back endsRun Coze alongside LangBot Models, Dify, n8n or Langflow and switch per pipelineCoze only
Extending behaviourPython plugins: commands, event listeners, tools, knowledge engines and moreCoze plugins and workflows inside Coze
ChannelsEvery platform LangBot supportsThe channels Coze offers for your site (coze.com or coze.cn)
Who controls the integrationYou, in your own LangBotCoze
Group-chat triggers and access controlConfigured per pipelineAs provided by each Coze channel
Self-hostingYesCoze is a hosted platform (Coze Studio is a separate open-source project)

What LangBot gives you out of the box

  • No deployment: cloud.langbot.app

    Sign up and build a bot in your browser. Free plan, public URL included.

    cloud.langbot.app
    1. Sign up
    2. Pick a platform
    3. Bot online
    Details ›

    Sign up at 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

    Top models ready to use. No API key to apply for.

    • ClaudeClaude
    • OpenAIGPT
    • GeminiGemini
    • DeepSeekDeepSeek
    • QwenQwen
    • ZhipuGLM
    • KimiKimi
    • GrokGrok
    No API key needed
    Details ›

    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

    Say what should change. Agents and the assistant carry it out after you approve.

    In the support group, only answer when someone @mentions the bot, and welcome new members.
    Routing: reply on @mention
    group.member_joined → welcome
    Applied after you confirm
    Details ›

    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 the goal; review the proposed changes; confirm.

    Proposed changes

    • Create pipeline “Support”
    • Attach knowledge base “FAQ”
    • Bind to the WeCom bot
    Confirm
    Details ›

    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

    One set of operations behind HTTP, MCP, CLI and webhooks; agents get MCP, Skills, a sandbox and knowledge bases.

    One Workspace API

    HTTP APIMCP ServerlbctlWebhooks

    Agent tools

    MCPSkillsSandboxRAGEvents
    Details ›

    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

    Seven component types in an isolated runtime, with permissions you grant.

    Plugin Runtime · isolated process

    RunnerToolCommandEventListenerKnowledgeEngineParserPageCertified Plugins
    Details ›

    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 lists plugins, MCP servers and Skills, with reviewed Certified Plugins.

  • Open source, with official backing

    Apache 2.0, official hosting and enterprise support.

    Apache 2.0Free commercial useOfficial CloudEnterprise Support
    Details ›

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

When Coze publishing is enough

If Coze already publishes to the channel you need and its behaviour fits, publishing directly is the shortest path.

When to add LangBot

  • The platform you need is not on Coze's channel list for your site.
  • You want one agent on many platforms with consistent group-chat rules and monitoring.
  • You want to switch the agent back end later (for example to Dify or your own model) without touching the bots.

Publish the agent with the API option, then enter the token, bot ID and API base URL in LangBot. See the Coze runner guide or Coze on Discord.

Further reading

FAQ

Does it work with both coze.com and coze.cn?

Yes. Use https://api.coze.com or https://api.coze.cn as the API base URL to match the site.