Adopt rtk-ai/rtk

[gjkim42]

https://github.com/rtk-ai/rtk
rtk will reduce the token usage 60-90%.

Adopting it in the framework level would be great.

[axon-agent]

🤖 Axon Agent @gjkim42

Axon Triage Report

Kind: kind/feature

Status: STILL VALID

Evidence:

• No merged PRs or commits reference rtk or this issue
• No existing rtk integration found in the codebase (grep for "rtk" returned zero results)
• rtk-ai/rtk is an actively maintained Rust CLI tool (v0.20.1, 1.1k stars) that acts as a proxy to reduce LLM token consumption by 60-90% by filtering and compressing command outputs (git, test runners, linters, etc.) before they reach the context window
• Axon runs agents (Claude Code, Codex, Gemini) inside Kubernetes Jobs. Integrating rtk at the framework level would mean installing it in the agent container images and configuring it as a shell wrapper or hook, so all tool-call outputs are compressed before reaching the LLM
• Related but distinct issues exist for token observability (API: Add cost/token Prometheus metrics and BudgetPolicy for production cost governance #355 — Prometheus metrics/BudgetPolicy, telemetry for real usage detection #390 — telemetry for usage detection), but nothing addresses token reduction at the tool-output level

Duplicates found: None found

Recommendation: KEEP OPEN — This is a valid feature request. rtk is a well-maintained, zero-dependency binary that could be bundled into Axon's agent images to reduce token consumption for all supported agents. However, the issue would benefit from more specifics: which agent types should be targeted first, whether this should be opt-in via AgentConfig, and how rtk's hook-based installation model maps to Axon's entrypoint flow.