Local proxy that compresses your LLM API requests so you pay less, with no change to the answers. Trims wasted tokens from prompts, history, tool output, and code before they're sent: -31% input / -74% output, measured live. Any provider, no extra model calls. Also an MCP server and embeddable library (Rust, Python, Ruby, Kotlin, Swift).
Insufficient evidence to grade. This server's source has not been statically analyzed, so a low grade would only mean "nothing found", not "nothing there". We don't show a reassuring grade we can't stand behind. Attested signals (CVEs, provenance) below still apply.
Once the source is analyzed (see the analysis flag in the header), a graded score appears here. How analysis works: methodology.
graded 12m ago · see ecosystem CVEs →
no known CVEs for this server.
No tool-safety findings — heuristic detectors run on the compute-risk cadence; a finding appears when a tool trips a rule.
[](https://mcpobservatory.com/servers/github:fkiene/llmtrim/security) Heuristic, inferred signals — false positives (legitimately powerful tools, forks, language ports) are expected. Treat each as "review this", not a verdict. See the ecosystem-wide picture on the security hub, or the fleet security of fkiene.