Gemini Spark
Consumer 24/7 personal agent in the Gemini app that runs in Google's cloud and acts across Gmail, Docs, Slides and connected apps.
Proprietary
SaaSLicense medium
Always-on AI Agents
Persistent personal and team agents that act on a user's behalf across apps, chat, and their own computer.
Always-on agents keep running after you close the chat: they hold their own identity, memory, and compute, and act across apps and messaging without a prompt each time. That autonomy is the risk — check each tool's approval gates, sandboxing, audit trail, and security record before allowing it on corporate data.
- In dataset
- 8
8 matching tools
More filters
Major vendor tools
Tools from the major platform vendors are listed first; independent and open-source options follow below.
Microsoft Autopilot
Microsoft
Always-on Copilot agent with its own Entra identity that works in Teams, Outlook and M365 on a user's behalf; built on OpenClaw.
Azure
Proprietary
SaaSLicense medium
Muse
Meta
Meta's consumer and small-business personal AI agent that keeps working on tasks in a secure VM and is reached via the Muse app or WhatsApp.
Proprietary
SaaSLicense medium
Filtered open source and third-party tools
5 matching non-vendor tools in this filtered view.
Grok Bot
xAI
xAI's always-on bots with their own cloud computers that work across apps 24/7; Team Bots add shared bots for Teams and Enterprise plans.
Proprietary
SaaSLicense medium
Hermes Agent
Nous Research
Self-improving open-source agent with persistent memory, built-in cron and a gateway to 20+ chat apps; runs on a laptop, VPS or cloud sandbox.
MIT
Self-hostedLicense low
Manus
Manus
Autonomous general AI agent with its own sandbox computer, scheduled tasks, Telegram access and a desktop 'My Computer' mode.
Proprietary
SaaSHybridSOC 2License medium
NanoClaw
NanoClaw
Lightweight OpenClaw alternative that runs each personal agent in its own container, messaged from WhatsApp, Telegram, Slack and more.
MIT
Self-hostedLicense low
OpenClaw
OpenClaw Foundation
Self-hosted personal AI agent that runs on your own machine and acts for you through WhatsApp, Telegram, Slack, Teams and 20+ chat apps.
MIT
Self-hostedLicense low
Important notes
Warning
Gemini Spark: Announced at Google I/O in May 2026: trusted-tester rollout, then a beta for US Google AI Ultra subscribers.
Warning
Grok Bot: Beta: launched 2026-08-11, expanded to more plans 2026-08-26; Team Bots in public beta on Teams and Enterprise plans since 2026-09-28; enterprise waitlist.
Warning
Manus: Manus announced it resumed independent operations on 2026-09-01; the transition involved data backup/restore and a temporary interruption to access.
Warning
Microsoft Autopilot: Private preview: announced as Microsoft Scout on 2026-06-02 for Frontier program organisations, renamed Autopilot on 2026-09-25 with private preview expanding at end of September 2026. Not Windows Autopilot (device provisioning).
Warning
Muse: Consumer launch 2026-09-08 in the US (iOS, Android, web); small-business version launched 2026-09-29 in the US and Canada.
Warning
OpenClaw: Security history: CVE-2026-25253 (1-click RCE via gateway token exfiltration, CVSS 8.8) affected versions <= 2026.1.28 and was fixed in 2026.1.29; all ClawHub skills have been scanned with VirusTotal since 2026-02-07.
Recent updates
See auto-detected changes2026-09-28
Grok Bot
Grok Bot
xAI opened Team Bots, shared always-on bots for Grok Teams and Enterprise plans, as a public beta.
Source2026-09-25
Microsoft Autopilot
Microsoft Autopilot
Microsoft renamed its always-on agent Microsoft Scout to Autopilot in the new Copilot, with usage-based billing and a private preview expanding to more Frontier customers at the end of September.
Source2026-09-08
Muse
Muse
Meta launched Muse in the US, a personal AI agent that runs tasks in its own secure VM and is reachable in the Muse app and WhatsApp.
SourceFrequently asked questions
What is an always-on AI agent?
An agent that keeps working after the chat closes: it holds its own memory, credentials or identity, and usually its own cloud or local computer, and acts across email, chat, files, and web apps on a schedule or trigger rather than one prompt at a time. Microsoft, Meta, xAI, and Google all announced or shipped products in this class in 2026, alongside open-source projects such as OpenClaw.
How is this different from an assistant or an agent framework?
Assistants answer when asked; agent frameworks are libraries developers use to build agents. Always-on agents are finished products that act on a person's or team's behalf without a prompt each time, which is why they need their own controls: approval gates for outbound actions, sandboxed compute, scoped identities, and an audit trail.
What should a security team check before allowing one?
Where it runs (vendor VM, user device, or self-hosted), what identity and credentials it holds, which actions need human approval, whether every action is logged to a system you control, and where third-party skills or plugins come from. Each per-tool page records the product's documented controls and any source-backed security incidents.
Explore adjacent hubs
Compare the active category against the platform foundation layer, the cross-category updates feed, and the sourcing/contribution guide.
Platforms (Platforms)
Review Microsoft Foundry, Amazon Bedrock, and Gemini Enterprise Agent Platform as the foundation layer behind this category.
Weekly updates (Updates)
Check the market-intelligence feed for high-impact moves, plus the expandable full log for releases, deprecations, acquisitions, and other notable changes.
About and contribution rules (About)
See sourcing standards, contribution rules, and project scope before adding or updating tracked tools.