Refactor messaging: emphasize transparency and control over configurability

[chinmaymk]

Summary

Updated marketing copy and documentation across the homepage, README, and concepts page to shift the messaging focus from "configurability" to "transparency and control." The new messaging emphasizes that ra is an agentic loop with no hidden abstractions, where every component is exposed and interceptable.

Key Changes

• Homepage hero section: Changed tagline from "One Loop. Infinite Agents" to "An agentic loop you can take apart and put back together" with emphasis on "nothing hidden behind abstractions"
• Feature cards: Rewrote all 6 feature descriptions to highlight:
  • No system prompt included by default (user-defined agents)
  • Multi-provider support (6 providers, same code)
  • Bidirectional MCP integration
  • Real context control with deterministic discovery
  • Middleware hooks at every step of the loop
  • Single binary with 4 interfaces
• README.md: Expanded the "What is ra?" section from 2 sentences to a detailed multi-paragraph explanation covering:
  • Core philosophy (transparency, no hidden abstractions)
  • Provider support and agent code portability
  • Built-in tools, sessions, memory, and skills
  • MCP bidirectional support
  • Context control capabilities
  • Multi-interface deployment options
  • Logging and tracing
  • Layered configuration system
• Concepts documentation: Mirrored the expanded README explanation with added internal documentation links

Notable Details

• Messaging now emphasizes what users can control rather than what the framework provides
• Removed the "infinite agents" framing in favor of "take apart and put back together"
• Shifted from "config-driven" language to "exposed via config and extended by writing scripts or TypeScript"
• Added specific technical details (FTS5, SQLite, bun build --compile, JSONL) to demonstrate substance behind the transparency claim

https://claude.ai/code/session_013cukumwqxk8cFFjefZ5nXK