San Francisco startup launches an intelligent agent platform: a persistent AI assistant on cloud-based virtual machines

Event Summary

On March 18, this San Francisco AI startup launched a personal agent platform at the “MuleRun Agentic Computer: Major Feature Update” event; on March 28, a recap of the event was uploaded to YouTube (the video is currently not public).

The core of the platform is to provide users with independent cloud-based virtual machines where agents run continuously, covering scenarios such as e-commerce, game development, and finance. Users can build agents using low-code tools like n8n or the coding framework LangChain, with the accompanying Creator Studio (currently in public beta) supporting one-click deployment and global payments.

Why This Matters

Most AI tools currently operate on a “ask one question, get one answer” model—after the conversation ends, the context is lost. What this platform aims to do differently:

• Transform from one-off Q&A to a long-running, continually accumulating, proactive workflow assistant;
• If it can truly be implemented stably, it may change business processes like inventory management and financial monitoring that require 24/7 operation.

Product Composition

• Cloud Virtual Machines: Each user gets an independent instance, allowing agents to run and schedule continuously.
• Building Methods:
  • Low-code: n8n
  • Coding Framework: LangChain
• Creator Studio (Beta): Currently in public beta, supports global payments, facilitating deployment and monetization.
• “Self-evolving” mechanism in three layers:
  1. Task Layer
  2. Domain Knowledge Layer
  3. Community Validation Layer The official statement claims that agents can gradually improve based on effective practices within user communities. VentureBeat’s report emphasizes this as a “proactive” rather than a passive response positioning; PR Newswire’s press release confirms the aforementioned features.

Risks and Uncertainties

• The March 28 video recap is not public, and some demonstrations cannot be independently verified.
• Terms like “self-evolving” have a marketing flavor, and actual effectiveness and generality remain to be seen.
• Current information mainly comes from the press conference and press releases, lacking reproducible empirical evidence.

Impact Assessment

Judgment Criteria:

• Clear Value Proposition: Ensures the continuity and operability of agents through resident virtual machines.
• Developer-Friendly: Compatible with n8n and LangChain, lowering the entry barrier.
• Insufficient Verification: The closed-loop demonstration and “self-evolving” effects require more time and public cases for validation.

References

• VentureBeat report: Focus on “Proactive” AI Narrative (link in original text)
• Official Blog: Creator Studio Beta highlights “Monetization in Minutes” (link in original text)
• PR Newswire: Key points from the press conference and standard PR information (link in original text)

Conclusion: The direction of “continuously operating agents” is still in its early stages; the most likely beneficiaries are developers and studios willing to quickly experiment and deploy products on the platform. For trading funds and secondary market participants, short-term relevance is low.