RPCS-1 Agent Tuner — Documentation

The RPCS-1 Agent Tuner gives developers a structural framework for configuring AI agents. Instead of debugging oscillation, overload, and freeze failures case-by-case, you describe your agent's environment and get parameter recommendations grounded in receiver dynamics.

How it works

Every recommendation flows through three steps:

1. Compute receiver primitives — your environment inputs are translated into five receiver primitives (TI, SG, FT, UE, AR) using the Matching Principle and basin stability geometry.
2. Map to platform parameters — the primitives are mapped to your target platform's parameter space (temperature, max_tokens, model, tool strategy, etc.).
3. Evaluate regime — the resulting profile is checked against the four stability boundaries (stable / near_oscillation / near_overload / near_freeze) and any warnings are surfaced.

All steps are deterministic. The same inputs always produce the same outputs.

Quick links

• Getting started — install the Python SDK
• The five primitives — TI, SG, FT, UE, AR explained
• Matching principle — Pred-09-5: TI ≈ 1/H
• Stability regimes — oscillation, overload, freeze
• Platform mappings — Anthropic, OpenAI, open source

Just want to try it?

The interactive tuner requires no installation and no account. Describe your agent and get recommendations in under 30 seconds.