Ruflo 3.64.0 released
A minor release adding a verified-outcome learning loop enabled by default, which trains the model on test and build results.
What changed in 3.64.0
- New learning loop records each turn and trains on verified outcomes from test, build, typecheck, or lint commands only.
- Learned lessons scoring 0.6 or higher are saved to
.claude-flow/data/learned-lessons.jsonand cited in prompts when matched. - Use
ruflo hooks verified-ingestor the MCP toolhooks_intelligence_verified-ingestto train SONA on verified turns. - Run
ruflo doctor --component learning-loopto check if outcomes are reaching SONA.
Check before you update
- ruflo-mods learning loop is on by default; set
learningLooptofalsein configuration to disable it.
Summarised by AgentGid from rUv's official release notes on GitHub, as reported there; read the original before relying on it. Price, status and release dates come from our daily data.
Ruflo at a glance
| What it is | Open-source agent meta-harness for Claude Code and Codex by rUv: it installs as Claude Code plugins (core tools, a visual console and mods that add routing and policy enforcement) or runs as an MCP server for Codex and … |
|---|---|
| Latest version | 3.80.0, Nov 21, 2026. This page is about an earlier release. |
| Release pace | 53 stable releases in the last 30 days; 3.64.0 came the same day as the previous one |
| Price | Free (OSS). Free and open source (MIT). Official pricing ↗ (checked Oct 11, 2026) |
| Downloads, last 7 days | 67.8K down 26.4% |
| Gid Score | 49 #47 in Agent orchestration and team workspaces |