~ $ axon.recall("context")
booting memory graph ... all regions online
synapses reached. memory returned.
context restored.
~ $ Sammy Junior_← back to the ecosystem

// part of the AXON ecosystem

rtkx

A token-optimising CLI proxy that compresses repetitive dev-command output before it reaches an LLM (100+ supported commands, a single Rust binary). rtkx is a fork of rtk-ai/rtkby Patrick Szymkowiak and contributors. It is the compression engine of the AXON context stack.

github.com/sammyjdev/rtkx· Apache-2.0.

// measured, not vibes

Not a better filter. A dependency AXON actually needs.

Fork vs. upstream: verified identicaldiff, 2026-07-09 · rtk-ai/rtk@develop v0.42.4
0 diffbyte-for-byte identical output on find, grep, and git log; checked directly, not assumed
3 commands testedsame corpus

// the fork doesn't compress better than the original. It never touched that code, so there's no reason to imply otherwise.

ccr: the feature upstream doesn't havesrc/axon/mcp/server.py:861,984,993 · 2026-07-09
load-bearingAXON's own "restore_context" MCP tool calls "rtkx ccr restore" directly: it does not run on upstream rtk
real round-trip tested56.3% smaller on disk

// one real file (docs/METRICS.md, 6,197 bytes) stored via ccr at 2,706 bytes, restored byte-for-byte. Storage compression, not a context-token claim.