I built OpenMolt: Think OpenClaw, but Fully Programmatic

Hi everyone 👋

I’ve been working on OpenMolt.dev, an open-source framework for building programmatic AI agents in Node.js.

The idea came from frustration with agent workflows that are mostly static or brittle. Many tools rely on fixed pipelines or prompt chains that break easily when something unexpected happens.

With OpenMolt, you define agents directly in code with:

• instructions
• tools
• integrations
• memory

The agent then runs a planning + execution loop, deciding which tools to use and how to complete the task. So instead of rigid workflows, the agent can adapt if something fails or if the situation changes.

Some features so far:

• planning/execution loop
• tool and API integrations
• memory (short-term and long-term)
• scheduling
• CLI runner
• capability-based permissions so agents only access the tools you explicitly allow

The goal is to make AI agents behave more like real software systems rather than prompt scripts.

I’d love feedback from other builders here:

• Does the idea make sense?
• What kinds of agents would you want to build with something like this?
• Are there features you’d expect in a framework like this?

Thanks!