log-compression

Compress logs for LLM analysis — Rust-powered, Python API. 40-60% token savings.

Cold storage that queries like hot storage. High-ratio JSONL compressor with block-level random access. Query cold storage in seconds — no restore, no full decompression. 5–13% ratio on JSONL logs, 25–57% smaller than gzip. 1-hour window via DuckDB in under 6s. Free for personal and open-source use.

MCP server that cuts Claude Code token usage: log compression (builds/tests/installs) + smart file reading (symbol-level source access via tree-sitter).

Log folding for coding agents: collapses a massive log to its distinct events with counts, hides nothing, and explains every fold as JSON. Ships its own agent skill for Claude Code, Codex and OpenCode (lessence --skill).

DuckDB extension to read PFC-JSONL compressed log files with block-level timestamp filtering

Crystal lang shard which provides a file-based Log::Backend that supports automatic log file rotation, compression, and purging

Convert any compressed JSONL archive to PFC cold storage — from S3, Azure, GCS or local disk. No intermediate files, no schema changes, no pipelines.

Token-optimized output for any CLI — your AI agent reads 12 lines instead of 800, raw log always archived. Adapters for pytest/jest/vitest/ruff/eslint/tsc + a compression ladder for everything else.

Session-log compressor & storage optimizer for Claude Code, Cursor, Codex, Gemini, Windsurf, Ollama, Aider. Protects transcripts, chat history, and SQLite stores.

Python interface for PFC-JSONL

Compute before you send: Rust evidence queries and an installable skill for Codex and Claude Code

Stream Fluent Bit logs directly to PFC-JSONL compressed archives (.pfc). Free for personal and open-source use.

Context compression for LLM prefill — logs, JSON, prose, code. 660K-param ML model, 95% reduction on structured data.

Codag — independent third-party profile of a public API surface, by API Evangelist. Codag is a Y Combinator (Summer 2026) developer-tools company building drop-in log compression for AI agents. It takes oversized infrastructure logs — from Kubernetes, Docker, AWS CloudWatch, Vercel, Railway, Datadog, Sentry, syslog and unstructured sources — and re

REM-sleep log compression for LLM agents — compress raw logs into lessons, purge the rest 💤

Turn large log files into a short, ranked summary your AI agent can read, then let it search the full logs on demand. Runs locally, removes secrets before anything reaches a model, and costs nothing to run. Works with Claude Code, Cursor, and other agents over the Model Context Protocol.