Skip to content
Agents tracked: 492 Releases (7d): 545 Agent status: 1 with issuesSince last limit reset: Claude 9d · Codex 7d For AI agents: Markdown Updated Nov 20, 2026

OpenClaw by OpenClaw Foundation (nonprofit); created by Peter Steinberger

Open-source personal AI assistant that runs on the user's own machine as a gateway and works through chat apps such as WhatsApp, Telegram, Slack, Discord and iMessage, with a skills system and swappable model plugins. It also ships native apps for macOS, iOS, Android, Windows and Linux. Its README warns that tools run on the host for the main session unless sandboxing is configured, and that inbound messages should be treated as untrusted input.

General-purpose AI agents Personal assistant agent

Use it in: Slack, Teams, Telegram CLI Desktop app Mobile app Works with: MCP Local models Own API key Multi-agent Headless / CI

WebsiteDocsGitHub

Checked Nov 5, 2026. Tap a verdict to see its source.How we checkedReport an error

What OpenClaw does

OpenClaw is a personal assistant agent from OpenClaw Foundation, first released in November 2025. You use it through messaging apps, a command-line tool, a desktop app and mobile apps on macOS, Linux, Windows, iOS and Android. Its documentation covers MCP servers, subagents or parallel agents, a non-interactive mode for scripts and CI, your own API key and local models. It can run models from Anthropic, OpenAI, Google, xAI, DeepSeek and Qwen and others. It is free and open source (MIT); you pay only for the models it calls.

Status: Renamed from Clawdbot to Moltbot on 2026-01-27 and to OpenClaw on 2026-01-30 (first published 2025-11-24 as Warelay), and stewarded by the OpenClaw Foundation after creator Peter Steinberger announced on 2026-02-14 that he was joining OpenAI.

Recent developments

  • 2026-02-14: creator Peter Steinberger announced he is joining OpenAI and the OpenClaw Foundation took over stewardship.
  • 2026-08-30: OpenClaw 2.0 (v2026.8.1) announced.

What people use it for

An assistant inside your chat apps

Ask it from WhatsApp or Telegram to sort your inbox, send an email or put a meeting in your calendar.

Talk to it from your phone

Connect a Telegram bot to the assistant running at home and message it while you are out.

First chat on your own computer

Install it, let the setup reuse the AI access you already have, and chat with it in a page in your browser.

One assistant shared by a team

Run one copy for a small team that trusts each other, so everyone reaches the same assistant from their chat.

Get help when setup breaks

Run its built-in check, which writes a summary of the problem and can pass it to a coding assistant on your computer.

Popularity
#1 in General
Gid Score 71 / 100 · #8 of 251 ranked AI agents
Trend
Slowing down
downloads down 24.3% over 30 days · up 5.8% last week
Community
395K GitHub stars
VS Code rating 3.5 / 5

Usage data: latest readings Nov 17 to Nov 19, 2026, last checked Nov 19, 06:32 UTC. Date of each number

OpenClaw popularity and usage data

OpenClaw was downloaded 3.1M times from npm in the week ending Nov 17, 2026, up 5.8% on the week before and down 24.3% over the last 30 days compared with the 30 before. The GitHub repository has 395K stars, 4,087 of them added in the last 30 days. The latest stable release is 2026.10.33, published Nov 18, 2026; there were 15 stable releases in the past 30 days. It was mentioned in 123 Hacker News posts and comments over the last 30 days.

Downloads per daynpm, 7-day average 445K/day
GitHub starsrunning total 395K
Hacker News mentionsposts and comments, rolling 7-day total 8 this week
Wikipedia page viewsEnglish Wikipedia, 7-day average 525/day
Website rankTranco list; a lower rank means more traffic #12,436

Download the raw daily series: openclaw.csv

Gid Score breakdown

Adoption78
Community100
Attention60
Momentum26

Gid Score shows public popularity and activity (usage, GitHub stars, mentions, growth), not the quality of the product. Each component is scored 0 to 100 from public signals; a dash means no data for it. How the Gid Score works.

How fresh these numbers are

NumberLatest readingLast checked
Package downloadsNov 17, 2026Nov 19, 06:22 UTC
GitHub starsNov 19, 2026Nov 19, 06:26 UTC
Homebrew installsNov 19, 2026Nov 19, 06:26 UTC
Docker pullsNov 19, 2026Nov 19, 06:26 UTC
App store ratingsNov 19, 2026Nov 19, 06:26 UTC
Website rank (Tranco)Nov 18, 2026Nov 19, 06:32 UTC
Wikipedia viewsNov 18, 2026Nov 19, 06:26 UTC
Hacker News mentionsNov 18, 2026Nov 19, 06:26 UTC

We fetch every source once a day. Registries publish downloads and page views a day or two late, so their latest reading trails the check. A reading more than 4 days old is marked out of date; the Gid Score still uses it until a newer one arrives.

Pricing

Free and open source (MIT) with no paid tier, hosted service or token; users pay their chosen model providers.

Source: official pricing page, checked Nov 5, 2026. Prices change often; confirm before you buy.

Compare prices of all agents

Quick answers

  • Free to start?Yes
  • Open source?Yes (MIT)
  • Bring your own API key?Yes
  • Runs onmacOS, Linux, Windows, iOS, Android
  • Where you use itMessaging apps, CLI, Desktop, Mobile
  • Supports MCP?Yes
  • Model providersanthropic, openai, google, xai, deepseek, qwen, mistral, moonshot, meta, amazon, openrouter, any OpenAI-compatible, local
  • Subagents / multi-agent?Yes
  • Runs in scripts / CI?Yes
  • Feature comparisonAll agents
  • Closest alternativeDeepSeek Harness · all alternatives

What OpenClaw users report right now

All 4 reports, questions and fixes →

Open Users report that on OpenClaw 2026.9.6 the Gateway's model-catalog worker keeps growing in memory, to several gigabytes and in some cases tens of gigabytes, until a critical memory-pressure event resets it, and then the cycle starts again. Some users saw stalls or lost service during the peaks; the issue is open.

Reported on GitHub · 3 👍

Open Users report that temporary plugin-capture folders under ~/.openclaw/tmp/plugin-captures are not cleaned up, so disk use keeps growing, by gigabytes per day on some hosts. A maintainer says a change that repeats the cleanup every hour was merged in October; the issue is still open.

Reported on GitHub · 2 👍

Open Users report that OpenClaw can silently stop loading MEMORY.md or USER.md into new sessions once it marks the file as untrusted, for example after the agent writes it in a turn that used web search or after the file was edited outside the agent. They say there is no warning and no supported command to restore it; the issue is open.

Reported on GitHub · 2 👍

Closed Users report that OpenClaw's built-in system agent failed with 'prepared model runtime plugin generation was superseded', often after a config reload, and then told them to run openclaw onboard even though normal model calls worked. The issue was closed by a merged change to that misleading advice; the thread does not show that every trigger of the failure was fixed.

Reported on GitHub · 1 👍

The most discussed problems in its GitHub repository, checked Nov 12, 2026. Written by AgentGid in its own words; statuses and fix versions come from the project's own pages. Reports are users' accounts, not AgentGid tests.

What changed: recent OpenClaw releases

37 stable releases in the last 90 days · Monthly summary · Full changelog

2026.10.33
  • Long conversations and large session stores use bounded export and recovery work, reducing out-of-memory failures during chat, memory export, Doctor repair, and Gateway startup. Authored
  • Interrupted updates recover when their executor dies instead of leaving the Gateway stuck offline, while version-skew checks protect state still used by an installed service. Authored
  • Replies survive more restart and reconnect paths, preserving original prompts, recovering Telegram responses, and avoiding lost WhatsApp replies or skipped Signal messages.
  • Shared-session privacy is stronger: Side chat stays private to its viewer, internal model prompts stay out of displayed output, and compaction redacts credentials in quoted summaries. Authored
  • Automations can react to polled conditions and MCP events, run repeatable scripts, and use explicitly allowed MCP servers. Authored
  • Installed skills can be searched and read on demand, helping agents find relevant instructions without putting every installed skill into the prompt. Authored
2026.11.5
  • Restart-safe conversations: Queued messages, send permissions, and final chat writes are retained across Gateway restarts. Authored
  • Safer updates: Failed Doctor or Gateway startup rolls back an activated update, while restored npm packages remain readable by non-root services. Thanks @RomneyDa, @obviyus, and @vincentkoc.
  • Telegram widgets: Interactive widgets now render in Telegram messages, including private-chat topics.. Related #197923 Thanks @obviyus and @dankarization.
  • Faster memory recovery: Large memory-index rebuilds finish reliably, and embedding indexing recovers when a gateway truncates a batch-limit error..
  • More reliable model fallback: A configured primary account no longer blocks Claude CLI fallback, and OpenAI sessions recover from compaction item conflicts.. Related #158421 Authored
  • Usable device folders: Existing device folders remain selectable with the default runtime, and each folder choice gets a clear, compact menu row. Thanks @RomneyDa and @obviyus.
Earlier releases
2026.11.4
  • Release notes — formatted for people.
  • Changelog — plain Markdown for AI agents and tools.
  • npm package
  • registry tarball
  • integrity: sha512-kSMzBc0fIWrW4DqneFEXL9FRboCZCQV22Edm6iFPqRaBexUeZp9huWNPVWLXmuSAQRezorPfrABkODqq65NOmw==
  • release SHA: d843d6af39a42cd3c30084edbeabeece732fd93a
2026.9.37
  • Private inspection stays private: paired-node photos and screenshots are no longer automatically attached to replies, and explicit media-read denials are honored when preparing reply attachments.
  • Responsive replies and pages: long zero-width Telegram table cells and malformed plugin SVG transforms no longer stall the Gateway or browser page.
  • Linked accounts survive resets: sessions keep their person-linked model account through /new, /reset, and idle or daily rollover. Fixes #179432. Thanks @masatohoshino Authored
  • Safer update recovery: interrupted snapshot writes preserve the previous rollback point, and profile-backed updates and Doctor maintenance no longer trip over their own maintenance ownership.
  • Connections recover cleanly: channel accounts can recover after a manual stop times out, and ClickClack reconnects after silent network drops. Fixes #184852.
  • Windows state access works across path aliases: equivalent Windows paths preserve the same state ownership instead of falsely reporting offline maintenance. Fixes #176163. Thanks @tjarrau-dev Authored
2026.11.3
  • Private task results: Keep private subagent completions out of streamed chat replies. Thanks @fuller-stack-dev.
  • Reliable child-task handoffs: Allow a parent to hand off to replacement work after cancelling a child, without leaving the cancelled child's completion receipt blocking the handoff. (#178893.
  • Visible tool failures: Report unanswered tool failures instead of silently ending a user-addressed turn as a success. (#182458; related #99645) Authored
  • Bounded authentication: Stop stalled provider authentication at the run deadline and propagate cancellation through authentication work. Thanks @amittell.
  • Sandboxed image access: Keep required-sandbox image previews inside allowed folders, even in Full Access mode, while preserving explicit administrator image allowances. Thanks @shakkernerd.
  • Verified workspace connections: Preserve saved node certificate pins when an empty fingerprint is supplied, and verify workspace-transfer certificates before sending credentials or data.
2026.11.2
  • Release notes — formatted for people.
  • Changelog — plain Markdown for AI agents and tools.
  • npm package
  • registry tarball
  • integrity: sha512-SgBml4WLp8V1DdAYkrTPLzVdA/iYo3iuHCtm4788GjaG83QZdd1VeOTlue0HBJGdn3SbFE06b6uPs+AM5ShMgg==
  • release SHA: b556fdf457b026574bf0f1e7b4914b15f3f179ce

Compare OpenClaw with another agent

vs
# Agent Gid Score Downloads 7d 7d 30d Stars Latest release Price
10 67 658K down 22.5% up 59.3% 284K+685/day — Free
13
Hermes AgentNous Research
63 88.2K up 34.8% — 265K+293/day 0.23.24d ago Free (OSS)
39
ManusButterfly Effect Pte. Ltd. (Manus AI), Singapore
51 — — — — — $20/mo
56
Cloudflare OSCloudflare
46 — — — 11.8K+10/day — Free (OSS)
59
Muse NewMeta
462/4 — — — — — Free
68
Andee NewKuafuAI
442/4 — — — 6,157+162/day 0.16.02d ago Free

Head to head: OpenClaw vs Hermes Agent · OpenClaw vs DeepSeek Harness · OpenClaw vs Manus · OpenClaw vs Cloudflare OS · OpenClaw vs AutoGPT · OpenClaw vs Genspark · OpenClaw vs nanobot · OpenClaw vs Octop · OpenClaw vs QwenPaw

Tell OpenClaw

“Every morning at 8, send me on Telegram a short list of today's calendar events and the emails that need a reply.”

  1. Install it on macOS, Linux or WSL2 with the one-line installer; it brings its own Node.js when needed and starts the setup wizard.

    curl -fsSL https://openclaw.ai/install.sh | bash

    From github.com/openclaw/openclaw/blob/main/README.md

  2. If you installed the package yourself with npm, run the setup wizard and install the background service.

    openclaw onboard --install-daemon

    From github.com/openclaw/openclaw/blob/main/README.md

  3. Open the web dashboard and send a first message.

    openclaw dashboard

    From docs.openclaw.ai/start/getting-started.md

Mistakes to avoid

Letting an assistant with shell, browser and web tools read emails and web pages on the host machine
Give untrusted content to a read-only reader agent, keep web and browser tools off for agents with tools, and turn on sandboxing

Instructions hidden in a page or email can steer the model into running tools or leaking context, even if you are the only sender.

Opening direct messages to everyone (dmPolicy open with "*") on a bot that has tools
Keep the default pairing mode or an allowlist and approve each sender

With pairing, a stranger's message never reaches the model; open DMs plus tools is one of the first findings the audit flags.

Installing any skill or plugin from the hub because it looks handy
Install only skills and plugins you trust, and set your own install policy

OpenClaw does not block dangerous code at install time; hub scans are not a full security check.

Making the Gateway reachable from the internet without checking the setup
Keep it on loopback or behind authenticated private access, and run openclaw security audit after config changes

The safe baseline keeps the Gateway local or authenticated; the audit tells you when you drift from it.

One shared Gateway for people who do not trust each other
Give each trust group its own Gateway and credentials, ideally separate OS users or hosts

The docs say one Gateway is one trust boundary, not a wall between adversarial users.

Get started with OpenClaw

npmnpm install -g openclaw
Homebrewbrew install openclaw-cli
Dockerdocker pull openclaw/openclaw

These commands use the packages tracked on this page. The vendor may recommend a different installer; see the official documentation.

OpenClaw FAQ

How much does OpenClaw cost?

Free and open source (MIT) with no paid tier, hosted service or token; users pay their chosen model providers.

Is OpenClaw open source?

Yes. OpenClaw is open source, released under the MIT license.

How popular is OpenClaw?

OpenClaw was downloaded 3.1M times from npm in the week ending Nov 17, 2026, up 5.8% on the week before and down 24.3% over the last 30 days compared with the 30 before. The GitHub repository has 395K stars, 4,087 of them added in the last 30 days. The latest stable release is 2026.10.33, published Nov 18, 2026; there were 15 stable releases in the past 30 days.

What is the latest version of OpenClaw?

The latest stable release we track is 2026.10.33, published on Nov 18, 2026.

What are the alternatives to OpenClaw?

The closest alternatives in the same category by Gid Score are DeepSeek Harness, Hermes Agent, Manus, Cloudflare OS.

More about OpenClaw

What's new

What's new in OpenClaw: November 2026 so far

Full monthly summary

November brought stability improvements across conversations, authentication, and data handling, with fixes for reliability issues in channel communications and workspace management. The month focused on preventing data loss and improving resilience under interruption.

  • Conversations now survive interruptions better, retaining history through oversized turns and preventing restarts when stopping early turns.
  • Safer Mac settings and workspace connections prevent secrets from being overwritten and verify certificates before sending credentials.
  • Channel replies stay reliable across Telegram, Discord, and webhooks even after request handlers or earlier turns finish.
  • Large conversations no longer exhaust worker memory during transcript trimming and Doctor repairs.

6 releases this month · summarised from the official release notes · all months

Articles

AgentGid articles about OpenClaw

All guides

Agent Skills
OpenClaw supports Agent Skills. Add ready-made skills for code review, slides, design, marketing and more: browse the Agent Skills catalog, with install commands for OpenClaw.
Used with it

Tools in projects built with OpenClaw

All tools

Most used libraries and frameworks in 326 public GitHub repos set up for OpenClaw, from their dependency files.

  1. TypeScript48.5%
  2. Vitest28.5%
  3. tsx23.0%
  4. Zod22.7%
  5. Playwright (JavaScript)21.2%
  6. dotenv19.6%
  7. ws16.6%
  8. yaml15.3%
  9. Express15.0%
  10. chalk14.4%

Used far more often here than in AI-built projects overall: what sets OpenClaw projects apart.

  1. grammY Transformer Throttler8.6% · 21.5×
  2. grammY Runner8.3% · 20.8×
  3. tslog10.1% · 20.2×
  4. node-edge-tts8.0% · 20.0×
  5. @lit/context9.8% · 19.6×
  6. Lit Signals9.8% · 19.6×
  7. signal-utils9.8% · 19.6×
  8. Pi Agent Core7.7% · 19.2×
Ratings

Ratings from people who use OpenClaw

No ratings yet. Used OpenClaw? Rate it and say what it is good and bad at.

News

OpenClaw news

All AI agent news

  1. Release GitHub Releases

    OpenClaw 2026.10.33 released

    Long conversations and large session stores use bounded export and recovery work, reducing out-of-memory failures during chat, memory export, Doctor repair, and Gateway startup. Authored; Interrupted updates recover when their executor dies instead of…

  2. Release GitHub Releases

    OpenClaw 2026.11.5 released

    Restart-safe conversations: Queued messages, send permissions, and final chat writes are retained across Gateway restarts. Authored; Safer updates: Failed Doctor or Gateway startup rolls back an activated update, while restored npm packages remain readable…

  3. Data AgentGid data

    OpenClaw downloads up 26% in a week

    3.1M npm downloads in the 7 days to Nov 12, against 2.5M the week before.

Price history

OpenClaw prices over time

Token prices of Anthropic modelsnewest Anthropic model in each family, output price per 1M tokens
Anthropic OpusAnthropic SonnetAnthropic Haiku
Sources
For the makers of OpenClaw badges, corrections

Add a live badge to your README or site:

AgentGid: #1 in General Gid Score: 71/100 downloads: 3.1M/week ▼24% 30d

[![AgentGid: #1 in General](https://agentgid.com/badges/openclaw-rank.svg)](https://agentgid.com/agents/openclaw/)
[![Gid Score: 71/100](https://agentgid.com/badges/openclaw-score.svg)](https://agentgid.com/agents/openclaw/)
[![downloads: 3.1M/week ▼24% 30d](https://agentgid.com/badges/openclaw-downloads.svg)](https://agentgid.com/agents/openclaw/)

Something wrong or outdated on this page? Send a correction with a link to the official source. More for makers.

Where these numbers come from

npm: openclaw. GitHub: openclaw/openclaw. Homebrew: openclaw-cli. Wikipedia: OpenClaw. Site rank: openclaw.ai on the Tranco list. See data sources for how each one is collected.

Product facts and pricing were checked against: github.com, raw.githubusercontent.com, en.wikipedia.org, registry.npmjs.org, formulae.brew.sh, hub.docker.com, itunes.apple.com, play.google.com.