The Short Answer
Kayba and Redapto both close the loop on agent improvement, but they operate at different levels. Kayba is a horizontal, open-source learning layer that works with any agent type. Redapto is a vertical solution focused specifically on customer support, auditing interactions to update SOPs and knowledge bases.
Kayba teaches any agent to learn from experience. Redapto optimizes customer support workflows.
What Each Tool Does
Redapto
Redapto (YC F25) builds self-improving systems for customer support:
- Interaction auditing: Reviews support conversations to identify gaps and failures
- SOP updates: Automatically revises standard operating procedures based on audit findings
- Knowledge base maintenance: Keeps support knowledge bases current with new patterns
- Support-specific: Purpose-built for customer support teams and their workflows
Redapto addresses a real pain point in support operations — SOPs go stale, knowledge bases drift, and agents repeat the same mistakes. Their approach is to automate the feedback loop within that specific domain.
Note: Redapto is early-stage with limited public technical detail. Their audit process does not appear to involve LLMs, relying instead on rule-based or traditional analysis methods.
Kayba
Kayba is an open-source learning layer (MIT, 2k+ stars) that synthesizes three published research papers into a unified framework:
- Recursive Reflector: REPL-based trace analysis that programmatically examines agent execution — grounded in the ACE framework (arXiv:2510.04618) and Reflective LLM Methods (arXiv:2512.24601)
- Skill extraction: Failures and successes are distilled into atomic, reusable skills with helpful/harmful counters
- Skillbook: A persistent, transparent collection of everything the agent has learned — organized, auditable, with provenance tracking back to source traces. Inspired by the Dynamic Cheatsheet approach (arXiv:2504.07952)
- Prompt generation: Approved skills are compiled into optimized system prompts
- Continuous learning: Delta updates refine the Skillbook incrementally as new traces come in
The framework is agent-agnostic and requires no fine-tuning — it works by improving the context your agent receives, not by retraining weights.
The Key Difference: Horizontal vs Vertical
Redapto is built for one domain: customer support. It speaks the language of that vertical — SOPs, knowledge bases, support interactions. If your problem is specifically "our support agents give inconsistent answers," Redapto is designed for that use case.
Kayba operates at the agent layer itself, independent of domain. The same framework that improves a coding agent also improves a support agent, a data analysis agent, or a browser automation agent. Instead of updating SOPs, Kayba extracts skills into a Skillbook and generates improved prompts — a mechanism that generalizes across any agent type.
| Aspect | Kayba | Redapto |
|---|---|---|
| Scope | Any agent, any domain | Customer support agents |
| Learning output | Skillbook entries + generated prompts | Updated SOPs + knowledge base entries |
| What improves | The agent's context (system prompts) | The support team's documentation |
Comparison
| Dimension | Kayba | Redapto |
|---|---|---|
| Open source | Yes, MIT license | No, closed-source |
| Domain scope | Horizontal — any agent type | Vertical — customer support only |
| Learning mechanism | Trace analysis, skill extraction, Skillbook, prompt generation | Interaction auditing, SOP/knowledge base updates |
| Research backing | 3 published papers (ACE, RLM, Dynamic Cheatsheet) | No published research |
| Audit approach | LLM-powered Recursive Reflector with REPL-based analysis | Non-LLM auditing (rule-based/traditional) |
| Human review | Built-in — approve, edit, or reject skills before deployment | Unclear from public information |
| Self-hosting | Yes, run entirely on your infrastructure | No, managed service |
| Framework dependency | Framework-agnostic (any agent, any trace format) | Support platform-specific |
| Pricing | Free (OSS) / $29/month (hosted dashboard) | Contact sales |
| Maturity | Production-ready, 2k+ GitHub stars, active community | Early-stage (YC F25) |
When to Choose Redapto
Redapto may be a fit if:
- Your problem is strictly customer support optimization
- You want SOP and knowledge base updates as the primary output
- You prefer a managed, support-specific solution over a general-purpose framework
- A non-LLM audit approach aligns with your compliance or cost requirements
When to Choose Kayba
Kayba is the stronger choice if:
- You run agents beyond customer support — coding, data analysis, browser automation, or multi-domain workflows
- You want a single learning framework across all your agents, not a point solution per vertical
- You need transparent, auditable improvements with full provenance tracking
- Self-hosting is a requirement (data sovereignty, air-gapped environments)
- You value open-source — inspect the code, contribute, fork if needed
- You want research-backed methods validated on public benchmarks (t2-bench: +27.4% pass@1, browser agents: 30% to 100% success rate)
Getting Started
Kayba is open-source and ready to use today:
pip install ace-framework
- Documentation — Setup guides and API reference
- GitHub — Source code and examples
- Dashboard — Hosted version with visual Skillbook management