# RubyLLM 2.x adds step-by-step agent loops, human-in-the-loop tool approval, and persistent agents

> SerpApi's blog covers the release of RubyLLM 2.0 and 2.1, which introduce incremental agent loop execution, human approval for tools, and database-backed agent persistence.

- Canonical: https://extractfeed.io/story/rubyllm-2-x-adds-step-by-step-agent-loops-human-in-the-loop-30be918/
- JSON: https://extractfeed.io/api/v1/stories/rubyllm-2-x-adds-step-by-step-agent-loops-human-in-the-loop-30be918.json
- Beat: Open Source & Tooling · Evidence: Primary source · Type/significance: release/2 · First seen: 2026-10-22T00:08:54.198113+00:00 · Updated: 2026-10-22T00:08:54.198113+00:00 · Edition: 2026-11-05
- Framing: model-written (headline, standfirst, why it matters, tags); source facts deterministic

## Briefing
- SerpApi: SerpApi's blog covers the release of RubyLLM 2.0 and 2.1, which introduce incremental agent loop execution, human approval for tools, and database-backed agent persistence.

## Why it matters
For Ruby developers building AI agents, these features address common pain points: debugging and controlling agent loops, adding a safety checkpoint for tool use, and surviving restarts without losing state. The persistence capability is particularly notable for production deployments where long-running agents need resilience. This signals a maturing Ruby ecosystem for agent development, though the impact is limited to RubyLLM users.

## Sources
- Primary source · SerpApi · 2026-10-21 — [RubyLLM 2.x: What's new for your agents](https://serpapi.com/blog/rubyllm-2-x-whats-new-for-your-agents/)

## Watch next
Will RubyLLM's persistence layer support distributed agent state across multiple processes or machines?

Topics: RubyLLM, SerpApi, ruby, agents, human-in-the-loop, persistence, agent-loop

---
extractfeed is an agent-readable changefeed for web scraping and data extraction. Index: https://extractfeed.io/agents.md