WiseClaw
Open source · Local-first · Multi-agent

Your dedicated AI researcherOnline around the clock

WiseClaw runs on your machine, browsing the web, watching your data sources, and executing tasks on a schedule — then delivers the findings straight to the channels you already use. No prompts to babysit, just goals to set.

License
MIT licensed
Channels
20+ delivery channels
Platforms
macOS · Windows · Linux

One interface, every agent in view

A visual control room puts setup, monitoring, and review just a few clicks away — power users can still drop into Advanced Mode.

Assign a task in plain language. WiseClaw breaks it into steps and keeps working, with every action fully traceable.

Placeholder mockup shown here — replace with real screenshots in the public/screenshots directory.

Built for continuous, autonomous operation

Eight core capabilities that take agents from reactive answers to proactive execution.

Autonomous browsing & monitoring

Agents visit pages, extract data, and track changes in the sources you care about — no manual checking required.

Multi-channel smart delivery

Findings and alerts land simultaneously across WhatsApp, Telegram, Slack, Discord, and 20+ other channels.

Local-first, fully in your control

Credentials and data stay on your device by default — the cloud is optional, never mandatory.

Scheduling & workflow orchestration

Chain multi-step tasks on a minute, hourly, or custom cadence, and let routine work run itself.

Multi-agent collaborative routing

Agents with different responsibilities can hand off work intelligently, coordinating on complex jobs.

Open, extensible skill ecosystem

Extend capabilities with pluggable Skills — freely mix and reuse contributions from the community.

Bring your own model

Works with Claude, GPT, Gemini and other leading models — switch or combine them per task.

Graphical UI, zero command line required

Everyday setup is fully visual out of the box, while a complete advanced mode remains for developers.

From install to steady output in four steps

No heavy operations required — configure once and let agents run autonomously for the long haul.

  1. 01

    Install & deploy

    Download the installer for your platform and get running locally in minutes — no server required.

  2. 02

    Connect sources & channels

    Point WiseClaw at the sites, APIs, or files you care about, and bind the channels you want results delivered to.

  3. 03

    Set the cadence

    Define how often each task runs and what triggers it, then hand it off to the agent.

  4. 04

    Keep receiving summaries

    Structured findings and alerts show up continuously in the chat apps you already use.

Ready to fit however you work

Four representative scenarios showing how WiseClaw slots into real workflows.

Financial analyst

Have an agent track industry filings, earnings, and sentiment continuously — with alerts on your phone the moment something moves.

Daily industry monitoring digest

3 relevant filings and 2 in-depth reports detected, ranked by importance with source links attached.

Chat tools vs. WiseClaw

The core difference: one waits for your prompt, the other keeps working for you.

Traditional chat toolsWiseClaw
Way of workingWaits passively for a promptSet a goal, then runs continuously
UptimeResponds only when askedRuns in the background, 24/7
Getting resultsYou have to open the chat to checkResults are pushed to your channels
Data ownershipMostly cloud-dependent processingLocal-first — data stays on your device
Task memorySingle-turn, easily loses contextPersistent memory with a traceable history
ExtensibilityFeature set is largely fixedOpen skill ecosystem, freely extensible

Local-first, with no vendor lock-in

A clean three-part architecture keeps data flow transparent and under your control.

Your device

Credentials, raw data, and execution history stay on your machine by default.

Model providers

Call Claude, GPT, Gemini and others on demand — swap or combine freely.

Delivery channels

Findings reach you through the channels you choose, with no extra third-party relay.

MIT licensed and open source — code and data flow are fully transparent
Cloud sync is optional; credentials and history are stored locally by default
Model providers are pluggable, so you're never locked into one vendor
You choose the delivery channel — messages don't pass through extra middlemen

What early users are saying

Real feedback from the private beta (shown with role-based attribution).

I used to refresh a dozen tabs a day by hand — now anomalies find me instead.

— An independent developer

Early beta user

Once we handed monitoring over to it, our stand-up prep time was cut in half.

— A startup team lead

Early beta user

Local-first matters a lot to us — we could adopt it without data ever leaving the device.

— A security practitioner

Early beta user

Coordinating tasks across multiple agents beats a single chat window by a wide margin.

— A product manager

Early beta user

Set it up once and forget it — the messages just show up in Slack on schedule.

— A data analyst

Early beta user

Open and extensible is why I picked it — the skill ecosystem is growing fast.

— An open-source contributor

Early beta user

These are placeholder quotes. Replace with real, permissioned user testimonials before launch.

Pick your platform and get started

Native installers for all three major platforms — up and running in minutes.

Version

v0.9.2

Download

macOS 12 Monterey or later · Apple Silicon / Intel

brew install --cask wiseclaw

Frequently asked questions

Can't find what you're looking for? Ask the community.

Web browsing and model calls need a connection, but your credentials and history stay on your device by default — cloud upload is never required.

20+ built-in channels including WhatsApp, Telegram, Slack, and Discord, plus custom Webhook support for anything else.

No. The graphical interface covers everyday setup end to end; the command line remains available as an advanced option for developers.

Claude, GPT, Gemini and other leading models are supported — switch or mix them freely depending on the task.

No. WiseClaw follows a local-first principle — credentials and runtime data are stored on your device by default.

Yes. You can create multiple agents with distinct responsibilities and let routing rules coordinate them on complex tasks.

WiseClaw uses an open-source Skill plugin system — build your own or install skill packs published by the community.

Fully open source under the MIT license — browse the code on GitHub and contribute anytime.