kimiclaw: Official Introduction

What kimiclaw Is

kimiclaw is presented as a configurable assistant rather than only a question-answering tool. It combines role customization, memory-oriented instruction, method learning, and schedule-driven execution so users can run long-term, repeatable workflows.

Within Kimi, kimiclaw is connected to OpenClaw, described as an AI assistant with a unique personality and long-term memory capabilities.

Deployment and Connection Paths

1) One-click deployment

• Users can create an OpenClaw instance from the Create Kimi Claw page.
• Kimi deploys OpenClaw to cloud resources with one click, without extra server purchase or command-line setup.
• Kimi automatically configures the Kimi K2.5 Thinking model and links Kimi Code membership quota, so no additional API setup is required.
• Kimi Web Search is enabled for web search support.
• One-click deployment is available on Allegretto plans and above, with details in Kimi membership plan information.

2) Link an existing OpenClaw

• If OpenClaw is already deployed, users can connect it in Kimi by installing the Kimi plugin.
• On the Kimi Claw page, choose "Link existing OpenClaw" and follow installation steps on the device running OpenClaw.
• After linking is complete, users can chat with that OpenClaw instance from Kimi.

Usage Framework for Consistent Results

Customize persona

Reply style is not fixed. A single instruction can change identity, tone, and output format. The guide suggests tuning three dimensions: name and role identity, speaking style, and fixed opening or closing patterns.

Use the built-in Skill Library

kimiclaw includes a Clawhub Skill Library so users do not always need to teach from scratch. The guide highlights information synthesis, analysis, and workflow execution as common scenarios.

• Information synthesis: news briefs, competitor comparison, meeting-note templates.
• Analysis: stock and industry mapping, data interpretation, risk extraction.
• Workflow: complete pipeline from requirements and decomposition to output and review.

Users can also teach private best practices by providing process, standards, and preferred output templates.

Set scheduled tasks

For stable task automation, define three elements together: time, output format, and constraints. This structure improves consistency and supports long-term automation habits.

At [time], do [task], output [format], and follow [constraints].

Sample requests include daily 9:00 market summaries (3 bullets plus 1 risk note), one-hour reminders for report continuation, and evening routines with a gentle tone.

Set up channels

kimiclaw can guide channel configuration with built-in Skills. For Telegram bot setup, users can provide token and pairing information in chat, and receive step-by-step guidance. Terminal UI support is noted as coming soon.

FAQ and Operational Notes

Can I control kimiclaw directly from terminal?

Not at present. Direct terminal access is not currently supported. A terminal UI for cloud workspace control is planned.

Is kimiclaw available on all platforms?

Current availability is web and mobile apps on iOS and Android.

What if kimiclaw does not reply?

• Refresh the page in case of server disconnection.
• Use "Restart kimiclaw" in Settings, then test messaging again.
• If needed, use "Auto-fix kimiclaw" in Settings.

Can kimiclaw send and receive files?

Users can send images and files to kimiclaw in Kimi for tool-based viewing. In the current version, sending files back from kimiclaw to users is not yet supported; users may use other channels to obtain needed files.

How do I uninstall the Kimi plugin in linked OpenClaw?

Use the command below only on linked OpenClaw devices. Do not run it on one-click deployed kimiclaw instances, because deleting that connection is unrecoverable.

bash <(curl -fsSL https://cdn.kimi.com/kimi-claw/uninstall.sh)

Entity Snapshot (SEO and Knowledge Clarity)

Editorial Scope and Accuracy Notes

This page is written in an official product-introduction style and is fully based on the provided Kimi Claw User Guide material. It focuses on currently stated capabilities, setup paths, usage methods, and FAQ behavior without adding unsupported claims.