NextClaw

English

简体中文

English

简体中文

Appearance

NextClawYour long-term personal AI partner

Install and start with built-in free-trial models—no API key required. NextClaw can read source material, use tools, work with local files, and create documents, charts, code, and apps on a machine you control.

Download NextClaw
Create your first task

Toward a Proactive Long-Term Partner ​

NextClaw's mission is to be a good long-term partner to you. The four stages toward 1.0 cover understanding you and your goals, proactive analysis and conversation, continuous self-improvement, and reliable long-term use. Initiative also means thinking independently and reaching out with something worth discussing, without requiring every exchange to become an execution task.

Full acceptance testing is not complete, and no release date is set. Read the vision · Explore the 1.0 feature roadmap. The guides below describe the current product.

What happens inside a task ​

Provide an outcome, source material, and a working directory. The agent can use models, skills, a browser, the terminal, and local files. You can observe, redirect, or stop it, then inspect the actual artifact in the same workspace.

1Describe the deliverable

State the material, output format, destination, and boundaries.

2Let the agent act

It reads files, invokes tools, runs commands, opens the web, or delegates subtasks.

3Stay in control

Watch actions, answer questions, add constraints, or stop execution.

4Inspect the result

Open replies, files, source, Markdown, HTML, images, or a Panel App.

A NextClaw task showing a data visualization beside a project document
Conversation, generated results, and project documents stay in one task.

Start with one real task ​

STEP ONEInstall and start your first taskThe built-in free trial needs no API key. You can switch to your own hosted or local model later.STEP TWOChoose a directory and outcomeBring real material and define the output, destination, and acceptance criteria.STEP THREEOpen the artifact and refine itReview the real file, chart, or page, then continue with the same context.

Keep the result beside the conversation ​

Open local files, Markdown, code, HTML, documents, spreadsheets, and presentations in the session workspace. Inspect HTML as source or rendered output and review code changes as diffs.

A local HTML data report open beside its NextClaw session
Review project files and a rendered data report without leaving the task.
Tasks and sessionsKeep goals, messages, actions, and follow-up changes together.Session workspaceOpen project files, subtasks, and session schedules.Results and previewsInspect documents, source, HTML, spreadsheets, and real files.Doc BrowserKeep websites, documentation, and browser tabs in the global dock.

Let background agents deliver finished work ​

Scheduled tasks, background agents, and long-running monitors can place reports in the AI Inbox. Read them when convenient, manage them with other results, or continue the conversation from the delivered report.

NextClaw Island theme showing a daily AI and technology briefing delivered to the AI Inbox
A delivered briefing stays beside conversations and apps in the same desktop workspace.

Turn repeated work into your own agents and apps ​

Dedicated agents can keep an identity, home directory, memory, skills, and runtime. Package proven work as a skill, connect external tools through MCP, or keep an interactive result as a Panel App.

Multiple dedicated agents in the NextClaw agent management page
Give different work its own agent.
AgentsSeparate roles, directories, skills, and runtimes.Skills and MCPPreserve working methods and connect more tools.Panel AppsUse dashboards, forms, and tools beside the task.Tools and actionsUnderstand file, terminal, web, message, and image actions.

Follow a complete task guide ​

Collect data and build a chart reportClean web or spreadsheet data and deliver charts, conclusions, and an HTML report.Organize a folder of local filesPreview the plan, then classify, rename, extract, and preserve a change record.Research multiple cited sourcesCompare web and local material and create a verifiable research document.Draft an article or reportVerify facts and structure before producing an editable draft.Prioritize customer feedbackMerge duplicate issues, count patterns, and preserve anonymous evidence.Generate and save an imageBegin with its use and composition, compare results, and refine locally.Build a small local appCreate HTML, a script, or a Panel App and debug it in the right side.Inspect and modify a codebaseReproduce, implement, run checks, and verify the real page.Generate and send a scheduled briefProve it manually, then choose the timing, directory, and channel.Start tasks from a chat appLet the local agent handle the request and send the result back.

Run work on schedule or from a messaging app ​

Scheduled tasks cover briefs, checks, and recurring summaries. Channels let requests and results move through Weixin, Feishu, Telegram, Slack, and other familiar entry points. Prove the work locally before automating it.

The NextClaw scheduled task management page
Review enabled state, previous runs, and the next schedule.
The NextClaw messaging channel configuration page
Connect familiar messaging apps to your agents.

Install, configure, and look things up ​

InstallationCompare the desktop app, npm, and Docker.Models and providersConnect hosted, local, or custom models.Security and permissionsReview directories, secrets, channels, and external tools.TroubleshootingStart with runtime status, logs, and connections.ConfigurationLook up complete settings and runtime options.Runtime resource usageReview the tested VPS baseline, idle memory measurements, and workload boundaries.CLI capability mapBrowse the complete NextClaw command surface by capability area.Product updatesSee what each version added, improved, and fixed.Project and communityExplore the vision, roadmap, activity, and community.

Released under the MIT License.

Copyright © 2026-present NextClaw