Recently browsing X, I saw many overseas bloggers recommending Kimi 2.5 for Clawdbot.
Pursuing high quality at a low price, people all over the world are the same.

Some time ago, many friends around me tested Kimi 2.5 and praised its excellent front-end aesthetics.
Until I scrolled to the post below, I was shocked!
Kimi 2.5 actually beat Gemini 3 Pro and Claude in the Design Arena rankings, and this is an open-source model.

It seems I missed quite a bit while researching Clawbot recently.
After some research, I quickly connected Kimi 2.5 to Clawdbot.
Below is a one-sentence test:
Write an exquisite Todolist webpage

The result is as follows:

Direct output from a single sentence, it's really impressive.
Front-end tasks are simple, so I tested a complex video generation Skill I wrote myself.
The process is long; it must first call Listenhub to generate audio and subtitle timelines.
Then generate image prompts based on subtitles, call Jimeng for image generation, then call Manim to generate transparent text animation, and generate the intro and outro based on my IP avatar.
Finally, use ffmpeg to stitch it all together into a complete video.

Previously, I only used Claude Opus 4.5; other models often had issues.
I didn't expect Kimi 2.5 to handle it too, which was unexpected.
Below is a detailed tutorial on how to connect Kimi 2.5 to Clawdbot (OpenClaw).
Also, I really have to complain that Clawdbot changed its name three times in one week:
Clawdbot -> Moltbot -> OpenClaw
I guess this is the last name change. This tutorial should be one of the most accurate on the web, haha!

To connect Kimi K2.5, first sign up for a Kimi membership plan.

After subscribing, go to the console to create an API key.
Note: The API key is only displayed once. Copy and save it in a safe place for later use.

Install OpenClaw
Taking a Mac as an example, open the terminal (Warp is recommended, and there seems to be a Windows version too).
Copy and paste the following command and press Enter.
curl -fsSL

After installation, enter the setup process.

Use the left and right arrow keys to select Yes and press Enter.
Select QuickStart for Onboarding mode.
Continue down, and later you will see Model/auth Provider.

Select Kimi Code API Key, as this is the plan we bought.

Paste the API key copied above and press Enter.

Select the first one to set it as the default model.

I skipped the bot configuration for Telegram and Discord; I will teach you how to configure Feishu later.
Then you reach Skill configuration; check them according to your needs.
Move with up and down arrows, select with space, and press Enter to confirm and proceed to the next step.

Then there is a series of configurations. If you don't have an API or don't need them for now, the three Hooks are all useful and recommended:
The first is to inject a Markdown file at startup, injecting content like a README at the start of a session.
The second is operation logging, recording commands and operation context executed in this session.
The third is to save the current session context summary when starting a new session for seamless transition.

If your computer is running a gateway or you have installed it before, it is recommended to select restart.

It asks where you want to use (incubate) your bot, with two options:
- TUI, which means Terminal UI, basically a command-line dialogue
- Web UI, which has a dialogue webpage

TUI is recommended by default, but I think Web UI might be more user-friendly for beginners.
They are not mutually exclusive; both will actually be installed.

If you want to enter TUI, type: openclaw tui

At this point, it's actually installed.
You can use Web UI or TUI for dialogues in the future.
However, the most interesting part of OpenClaw is its support for many IM tool integrations.
Talking to OpenClaw and giving commands through familiar chat tools feels like being a leader.
Below is the introduction to Feishu and other integration methods.
Connecting Feishu to OpenClaw
Someone developed a Feishu plugin for OpenClaw:
Copy the following command, paste it into the terminal, and press Enter.
openclaw plugins install @m1heng-clawd/feishu
After installation, the following interface will be displayed.

Open the Feishu Open Platform:
Click "Create Custom App," fill in the app name and description.

Then go to "Add App Capabilities" -> find Bot, and click "Add." (Some permissions require bot capabilities first)

Copy the App ID and App Secret on the app's "Credentials & Basic Info" page.

In the terminal, enter:
openclaw config set channels.feishu.appId "Your App ID" openclaw config set channels.feishu.appSecret "Your App Secret" openclaw config set channels.feishu.enabled true

Then you need to restart the gateway. Enter the following command in the terminal and press Enter:
openclaw gateway restart
The above operation is very important, otherwise, an error will occur when configuring events and callbacks in Feishu (connection not established).
Go back to the Feishu app's "Permission Management" page, click to enable permissions, and enter im:message.
Continue searching for keywords and enable the following app identity permissions:
- contact:user.base:readonly
- im:message
- im:message.p2p_msg:readonly (requires bot capability)
- im:message.group_at_msg:readonly (requires bot capability)
- im:message:send_as_bot
- im:resource (upload images or file resources)
Here comes the key point where many people make mistakes.
In event configuration and callback configuration, select "Long Connection" for the subscription method.
Then, in event configuration, click "Add Event" and add the following:
- im.message.receive_v1 (required)
- im.message.message_read_v1
- im.chat.member.bot.added_v1
- im.chat.member.bot.deleted_v1
After configuration, create a version and publish it on the "Version Management & Release" page.
Now open Feishu and search for "OpenClaw" to find the app bot.
Don't be intimidated by the number of screenshots; it actually only takes a few minutes to complete.
There is also a lazy configuration method, which is to let AI help you (method from stin):
"Help me connect Feishu to openclaw, plugin installation command: openclaw plugins install @m1heng-clawd/feishu"
However, after the plugin is installed, you need to enter the following command in the terminal to restart the gateway:
openclaw gateway restart
Then, send the Feishu App ID and Secret to it to complete the configuration.
(Note: You still need to manually configure the Feishu Open Platform yourself)
One highlight of OpenClaw is its support for various IM software.
Common ones include TG, Discord, WhatsApp, iMessage, etc.
Personally, I feel Feishu is the most user-friendly for Chinese users, TG is the simplest, and Discord is slightly more complex to configure but has the best display format.
A Discord bot configuration tutorial written by a friend:
Common Pitfalls in Installation and Use
I read a tutorial written by my friend YC and found that I have encountered many of them too.
I recommend everyone remember a few common commands:
Start OpenClaw's TUI:
openclaw tui
Restart the gateway:
openclaw gateway restart
Start a new conversation:
/new
Add a fallback model:
openclaw models fallbacks add [model company code/model name]
Example: openai-codex/gpt-5.2-codex
Set the default model:
openclaw models set [model company code/model name]
Example: kimi-code/kimi-for-coding
Also, since you are already connected to OpenClaw, you can ask it anything you don't know.
For example, I asked how to specify different models for different session threads; there are actually quite a few scenarios.
A Few Thoughts
I found that many people install OpenClaw but don't know what to use it for.
I think it's a lack of scenarios and exclusive Skills.
OpenClaw automatically loads Skills installed globally by Claude, which is quite convenient.
If you don't know how to write Skills, you can start by imitating or using others' Skills.
Sharing a curated OpenClaw Skill library with over 700 Skills collected.
Closing Remarks
This might be one of the most detailed tutorials I've written.
But this is just the beginning.
OpenClaw's gameplay goes far beyond this: long-term memory, heartbeat inquiries, asynchronous work, scheduled scripts...
Also, I haven't had time to study Kimi K2.5's multi-Agent cluster.
I feel it fits well with OpenClaw's philosophy.
However, tinkering is tinkering.
But always remember: The value of a tool lies in it actually being used.
I hope this tutorial can help friends who want to use OpenClaw in depth.
If you think this article is good, please support Qiao Bangzhu with a like, share, and follow.





