[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_126":127},"routes/modules.$creator.$moduleId",{"_14":15,"_16":17,"_18":19,"_20":21},"routes/modules.$creator.$moduleId.endpoints",{"_11":12,"_13":-5},"error","Module not found","tools","module",{"_20":22,"_23":21,"_24":25,"_26":27,"_28":29,"_30":31,"_32":33,"_34":25,"_35":36,"_37":-5,"_38":-5,"_39":-5,"_40":41,"_42":43,"_44":45,"_46":45,"_47":48,"_49":50,"_51":52,"_53":-5,"_54":45,"_55":56,"_57":58,"_59":-5,"_60":-5,"_61":-5,"_62":-5,"_63":64,"_65":66,"_67":68,"_69":70,"_71":72,"_73":-5,"_74":-5,"_75":-5,"_76":45,"_77":-5,"_78":79,"_80":81,"_82":83,"_84":85,"_86":87,"_88":89,"_90":91},"apiUrl","https://odel.app/api","creator","ojuschugh1","moduleId","sqz","@ojuschugh1/sqz","displayName","tagline","Pre-injection context compression for coding agents. Zero LLM calls, zero telemetry, offline-safe.","iconKey","https://avatars.githubusercontent.com/u/79078267?v=4","category","local","status","published","visibility","public","metaDescription","deploymentType","external","workerName","bindingName","mcpConfig","pricingModel","free","creatorUserId","system:external-import","createdAt",1787549931000,"updatedAt","sourceRegistry","official-mcp-registry","sourceUrl","https://registry.modelcontextprotocol.io/v0/servers?search=io.github.ojuschugh1%2Fsqz","sourceVerified",true,"sourceMissingSince","importedAt","upstreamStars",609,"upstreamUpdatedAt",1787500453000,"detectedTransport","detectedAuthType","probeStatus","lastProbedAt","githubRepoUrl","https://github.com/ojuschugh1/sqz","githubRepoFullName","ojuschugh1/sqz","releaseId","ext-ojuschugh1-sqz","version","1.6.1","descriptionMarkdown","

\n  ███████╗ ██████╗ ███████╗\n  ██╔════╝██╔═══██╗╚══███╔╝\n  ███████╗██║   ██║  ███╔╝\n  ╚════██║██║▄▄ ██║ ███╔╝\n  ███████║╚██████╔╝███████╗\n  ╚══════╝ ╚══▀▀═╝ ╚══════╝\n

\n\n

\n Compress LLM context to save tokens and reduce costs\n

\n\n

\n _{\n Real session stats:\n 3,003 compressions ·\n 178,442 tokens saved ·\n 24.7% avg reduction · up to\n 92% with dedup\n}\n

\n\n

\n $\"Featured\"$ \n

\n\n

\n $\"Crates.io\"$ \n $\"npm\"$ \n $\"PyPI\"$ \n $\"VS$ \n $\"Firefox\"$ \n $\"JetBrains\"$ \n $\"Discord\"$ \n $\"Homebrew\"$ \n

\n\n

\n Install ·\n How It Works ·\n Supported Tools ·\n Changelog ·\n Discord\n

\n\n---\n\nsqz compresses command output before it reaches your LLM. Single Rust binary, zero config.\n\nThe real win is dedup: when the same file gets read 5 times in a session, sqz sends it once and returns a 13-token reference for every repeat.\n\n```\nWithout sqz: With sqz:\n\nFile read #1: 2,000 tokens File read #1: ~800 tokens (compressed)\nFile read #2: 2,000 tokens File read #2: ~13 tokens (dedup ref)\nFile read #3: 2,000 tokens File read #3: ~13 tokens (dedup ref)\n─────────────────────── ───────────────────────\nTotal: 6,000 tokens Total: ~826 tokens (86% saved)\n```\n\n> [!NOTE]\n> **Name disambiguation:** this repo, [ojuschugh1/sqz](https://github.com/ojuschugh1/sqz), is an independent project and is not affiliated with any other similarly named or working tool or other compression projects that shorten \"squeeze\". If you installed `sqz` / `sqz-cli` / `sqz-mcp` from crates.io, npm, PyPI, or Homebrew, it comes from this repository.\n\n## Token Savings\n\n> **24.7%** average reduction across 3,003 real compressions ·\n> **92%** saved on repeated file reads ·\n> **86%** on shell/git output ·\n> **13-token** refs for cached content\n\nOne developer's week, measured from actual `sqz gain` output:\n\n```\n$ sqz gain\nsqz token savings (last 7 days)\n──────────────────────────────────────────────────\n 04-13 │ │ 2,329 saved\n 04-14 │ │ 0 saved\n 04-15 │███ │ 12,954 saved\n 04-16 │██ │ 9,223 saved\n 04-17 │████ │ 14,752 saved\n 04-18 │██████████████████████████████│ 105,569 saved\n 04-19 │████████ │ 30,882 saved\n 04-20 │█ │ 4,334 saved\n──────────────────────────────────────────────────\n Total: 3,003 compressions, 178,442 tokens saved (24.7% avg reduction)\n```\n\n### Per-command compression\n\nSingle-command compression (measured via `cargo test -p sqz-engine benchmarks`):\n\n| Content | Before | After | Saved |\n|---|---:|---:|---:|\n| Repeated log lines | 148 | 62 | **58%** |\n| Large JSON array | 259 | 142 | **45%** |\n| JSON API response | 64 | 53 | **17%** |\n| Git diff | 61 | 54 | **12%** |\n| Prose/docs | 124 | 121 | **2%** |\n| Stack trace (safe mode) | 82 | 82 | **0%** |\n\n### Session-level with dedup\n\nWhere the real savings live — the cache sends each file once, repeats cost 13 tokens:\n\n| Scenario | Without sqz | With sqz | Saved |\n|---|---:|---:|---:|\n| Same file read 5× | 10,000 | 826 | **92%** |\n| Same JSON response 3× | 192 | 79 | **59%** |\n| Test-fix-test cycle (3 runs) | 15,000 | 5,186 | **65%** |\n\nSingle-command compression ranges from 2–58% depending on content. Repeated reads drop to 13 tokens each. Your mileage will vary with how repetitive your tool calls are — agentic sessions with many file re-reads see the biggest wins.\n\n## Install\n\n**Prebuilt binaries** (no compiler required — works on every platform):\n\n```sh\n# macOS / Linux\ncurl -fsSL https://raw.githubusercontent.com/ojuschugh1/sqz/main/install.sh | sh\n\n# Windows (PowerShell)\nirm https://raw.githubusercontent.com/ojuschugh1/sqz/main/install.ps1 | iex\n\n# Any platform via npm\nnpm install -g sqz-cli\n\n# macOS / Linux via Homebrew\nbrew tap ojuschugh1/sqz\nbrew install sqz\n```\n\n**Build from source via Cargo:**\n\n```sh\ncargo install sqz-cli sqz-mcp\n```\n\n`sqz-cli` provides the `sqz` binary; `sqz-mcp` provides the MCP server. `sqz-engine` is a library dependency — it compiles automatically and does not need to be installed separately.\n\n**Build from source** (`cargo install sqz-cli`) works too, but needs a C toolchain:\n\n- Linux: `build-essential` (apt) or equivalent\n- macOS: Xcode Command Line Tools (`xcode-select --install`)\n- **Windows: Visual Studio Build Tools with the \"Desktop development with C++\" workload.** Without these, `cargo install` fails with `linker link.exe not found`. If you don't already have them, use the PowerShell or npm install above instead.\n\nThen initialize:\n\n```sh\nsqz init --global # hooks apply to every project on this machine\n# or\nsqz init # hooks apply to just this project (.claude/settings.local.json)\n```\n\n`--global` writes to `~/.claude/settings.json` (the user scope per the\n[Anthropic scope table](https://docs.claude.com/en/docs/claude-code/settings)),\nso the sqz hook fires in every Claude Code session on this machine. This is\nthe common case on first install. Your existing `permissions`, `env`,\n`statusLine`, and unrelated hooks in `~/.claude/settings.json` are\npreserved — sqz merges its entries rather than overwriting.\n\nPlain `sqz init` (project scope) is useful when you want sqz active only\ninside one repo.\n\n**Only using one agent?** Pass `--only` (or `--skip`) to limit which\nconfigs are written:\n\n```sh\nsqz init --only opencode # just OpenCode, nothing else\nsqz init --only opencode,codex # OpenCode and Codex\nsqz init --skip cursor,windsurf # everything except Cursor and Windsurf\n```\n\nAccepted names: `claude`, `cursor`, `windsurf`, `cline`, `gemini`,\n`kiro`, `opencode`, `codex`. Aliases (`claude-code`, `gemini-cli`, `roo`,\n`kiro-cli`) also work. `--only` and `--skip` can't be combined.\n\n### Manual installation (preserve comments in your config)\n\n`sqz init` round-trips your config file through a JSON parser to merge\nthe sqz entry, which drops any comments in your `opencode.jsonc` (and\nthe analogous JSON-with-comments files other tools accept). If you've\ncommented your config carefully and want to keep them, install by hand\ninstead.\n\n**OpenCode** — two steps:\n\n1. Drop the plugin file in place. `sqz` prints the generated TS to\n stdout so you don't have to hand-write the path-escaping logic:\n\n ```sh\n mkdir -p ~/.config/opencode/plugins\n sqz print-opencode-plugin > ~/.config/opencode/plugins/sqz.ts\n ```\n\n2. Add the MCP entry to your existing `opencode.jsonc` yourself.\n Append this block inside the top-level `mcp` object (create the\n `mcp` object if it doesn't exist):\n\n ```jsonc\n \"sqz\": {\n \"type\": \"local\",\n \"command\": [\"sqz-mcp\", \"--transport\", \"stdio\"],\n \"enabled\": true\n }\n ```\n\nComments in the rest of your file stay put. OpenCode auto-discovers\nthe plugin file; no `plugin` array entry needed (adding one causes\ndouble-loading, see issue #10).\n\n**Other tools** — Claude Code, Cursor, Windsurf, Cline, Gemini CLI,\nand Codex use plain JSON configs without comment support, so the\nautomated path is non-destructive there. Use `sqz init --only `\nfor those.\n\nThat's it. Shell hooks installed, AI tool hooks configured.\n\n## How It Works\n\n

\n $\"sqz$ \n

\n\nsqz installs a PreToolUse hook that intercepts bash commands before your AI tool runs them. The output gets compressed transparently — the AI tool never knows.\n\n```\nClaude → git status → [sqz hook rewrites] → compressed output (85% smaller)\n```\n\nWhat gets compressed:\n- **Shell output** — 40+ per-command formatters (git, cargo, npm/pnpm/yarn, pytest, ruff, go test, docker, kubectl, aws, terraform, gradle, gh, grep/rg, tree, curl, and more)\n- **JSON** — strips nulls, compact encoding, TOON format\n- **Logs** — collapses repeated lines\n- **Test output** — shows failures only (state-machine parsers for Rust, Go, Python, JS, JVM)\n\nWhat doesn't get compressed:\n- Stack traces, error messages, secrets — routed to safe mode (0% compression)\n- Your prompts and the AI's responses — controlled by the AI tool, not sqz\n\n## Supported Tools\n\n| Tool | Integration | Setup |\n|---|---|---|\n| Claude Code | PreToolUse hook (transparent) | `sqz init` |\n| Cursor | PreToolUse hook (transparent) | `sqz init` |\n| Windsurf | PreToolUse hook (transparent) | `sqz init` |\n| Cline | PreToolUse hook (transparent) | `sqz init` |\n| Gemini CLI | BeforeTool hook (transparent) | `sqz init` |\n| Kiro | Steering + MCP server | `sqz init` |\n| OpenCode | TypeScript plugin (transparent) | `sqz init` |\n| Codex CLI | AGENTS.md guidance + MCP server | `sqz init` |\n| Zed | AGENTS.md guidance + MCP server | `sqz init` |\n| Copilot CLI | preToolUse hook (transparent) | `sqz init` |\n| Copilot coding agent (CI) | Repo-level hook, self-bootstraps in the cloud sandbox | `sqz init --ci` + commit |\n| Any MCP server | `sqz-mcp proxy` wraps it, compresses its tool results | see below |\n| VS Code | [Extension](https://marketplace.visualstudio.com/items?itemName=ojuschugh1.sqz) | Install from Marketplace |\n| JetBrains | [Plugin](https://plugins.jetbrains.com/plugin/31240-sqz--context-intelligence/) | Install from Marketplace |\n| Chrome | Browser extension | ChatGPT, Claude.ai, Gemini, Grok, Perplexity |\n| [Firefox](https://addons.mozilla.org/en-US/firefox/addon/sqz-context-compression/) | Browser extension | Same sites |\n\n## Compress Any MCP Server\n\n`sqz-mcp proxy` sits between your agent and any stdio MCP server and compresses\nwhat flows back: tool results go through the full sqz pipeline (dedup refs on\nrepeats, safe-mode for stack traces and secrets, error results untouched), and\nverbose tool descriptions get compacted so `tools/list` stops eating your\ncontext window. Everything stays reversible — the proxy injects an `sqz_expand`\ntool so the agent can recover any original byte-exact.\n\nWrap a server by prefixing its command in your MCP config:\n\n```json\n{\n \"mcpServers\": {\n \"github\": {\n \"command\": \"sqz-mcp\",\n \"args\": [\"proxy\", \"--\", \"npx\", \"-y\", \"@modelcontextprotocol/server-github\"]\n }\n }\n}\n```\n\nFlags: `--lazy-tools` shortens every tool description to one sentence and\ninjects an `sqz_tool_help` tool that serves the full original docs on demand\n(some MCP servers spend 10-40k tokens on `tools/list` alone); `--no-desc`\nkeeps descriptions verbatim; `--no-cache` disables dedup refs. Works with\nevery MCP client (Claude Code, Cursor, Windsurf, Zed, Codex, Kiro, ...)\nbecause the client just sees a normal MCP server.\n\n## CLI\n\n```sh\nsqz init --global # Install hooks for every project on this machine\nsqz init # Install hooks for just this project\nsqz init --only kiro # Only configure Kiro (skip the rest)\nsqz init --only opencode # Only configure OpenCode (skip the rest)\nsqz init --skip cursor # Configure every agent except Cursor\nsqz init --ci # Also write .github/hooks/sqz.json for Copilot agent CI runs\nsqz compress # Compress (or pipe from stdin)\nsqz compress --no-cache # Compress without dedup (always full output)\nsqz expand # Recover original content from a §ref:HASH§ token\nsqz compact # Evict stale context to free tokens\nsqz reset # Clear dedup cache or compression stats\nsqz gain # Show daily token savings (bar chart)\nsqz gain --project . # Per-project daily gains\nsqz gain --days 30 # Last 30 days\nsqz stats # Cumulative compression report (includes regret signals)\nsqz stats --cost # Estimated $ saved under a prompt-cached billing model\nsqz stats --breakdown # Per-command token usage breakdown\nsqz stats --project . # Stats for current project only\nsqz stats --project list # List all tracked projects\nsqz discover # Find missed savings\nsqz resume # Re-inject session context after compaction\nsqz vizit # Live terminal dashboard (like htop for AI agents)\nsqz hook claude # Process a PreToolUse hook (Claude Code)\nsqz hook kiro # Legacy; Kiro now uses steering + MCP (sqz init)\nsqz print-opencode-plugin # Print OpenCode plugin TS for manual install\nsqz proxy --port 8080 # API proxy (compresses full request payloads)\n```\n\n### Dedup Escape Hatch\n\nWhen sqz sees the same content twice, it returns a compact `§ref:HASH§` token\ninstead of the full text. Most models handle this fine, but some (e.g., GLM 5.1)\ncan't parse the ref format and loop. Four ways to work around this:\n\n```sh\n# 1. Recover original content from a ref\nsqz expand a1b2c3d4 # prefix match\nsqz expand '§ref:a1b2c3d4§' # paste the whole token\n\n# 2. Compress without dedup (per-invocation)\necho \"...\" | sqz compress --no-cache\n\n# 3. Disable dedup globally (env var)\nexport SQZ_NO_DEDUP=1\n\n# 4. MCP passthrough tool (returns input byte-exact, zero transforms)\n# Available via tools/list when sqz-mcp is running\n```\n\n## Track Your Own Savings\n\nRun `sqz gain` in your shell any time to see your own daily breakdown (see the\nToken Savings section above for what the output looks like), and `sqz stats`\nfor the full cumulative report:\n\n```sh\n$ sqz stats\n 📊 sqz compression stats\n ──────────────────────────────────────────────────\n\n 178,442 tokens saved\n ↓ 24.7% average reduction\n\n Compressions 3,003\n Tokens in 721,840\n Tokens out 543,398\n Tokens saved 178,442\n Avg reduction 24.7%\n\n 🗄️ Cache\n ──────────────────────────────────────────────────\n Entries 43\n Size 39.1 KB\n```\n\nAdd `--breakdown` to see exactly which commands consume the most tokens:\n\n```sh\n$ sqz stats --breakdown\n\n 🔍 Top Token Consumers\n ──────────────────────────────────────────────────────────────────────\n command calls tokens in out saved\n ──────────────────────────────────────────────────────────────────────\n dedup 249 45541 3237 93%\n stdin 51 30851 24289 21%\n auto 132 18288 7740 58%\n echo 17 1050 558 47%\n ls -la 8 948 948 0%\n cargo build 7 170 145 15%\n git status 4 56 8 86%\n ──────────────────────────────────────────────────────────────────────\n```\n\n### Honest accounting: cost and regret\n\nToken reduction is not automatically billed-cost reduction — with provider\nprompt caching, most context re-transmits at a ~90% discount, and compression\nthat drops something the agent needed costs extra turns instead\n([arXiv:2607.12161](https://arxiv.org/abs/2607.12161)). sqz measures both\nsides instead of hand-waving:\n\n- `sqz stats --cost` estimates dollars saved under a prompt-cached billing\n model (cache write ×1.25 once, cache read ×0.10 per subsequent turn), with\n every assumption printed and overridable (`--price-in`, `--reread-turns`,\n `--cache-write-mult`, `--cache-read-mult`). This is the conservative\n estimate: without caching the same savings would bill at full input price\n on every turn.\n- `sqz stats` reports regret signals: quick re-runs (the agent re-produced\n byte-identical output within 2 minutes — the repeat bought nothing) and\n ref expands (`§ref§` tokens recovered to original bytes). Both are proxies\n for \"compression dropped something the model needed.\" If one command\n keeps showing up, its formatter needs work — file an issue.\n\nsqz's design already avoids the failure modes that study measured:\ncompression is deterministic and query-agnostic (never invalidates provider\nprompt caches), only new tool output is touched (history is never rewritten),\npiped or redirected commands are never intercepted, and the 16-token net-win\ngate skips compressions that wouldn't pay for their own markers.\n $\"image\"$ \n\n\n**Per-project filtering:**\n\n```sh\nsqz stats --project . # stats for current project only\nsqz stats --project list # list all tracked projects\nsqz gain --project . # daily gains for current project\nsqz gain --days 30 # last 30 days instead of 7\nsqz gain --days 30 --project . # combine both\n```\n\nStats are stored locally in SQLite under `~/.sqz/sessions.db` — nothing leaves your machine.\n\n## How Compression Works\n\n1. **Per-command formatters** — 40+ commands across 9 ecosystems get purpose-built compression:\n\n | Ecosystem | Commands |\n |---|---|\n | Git | status, log, diff, show, stash, remote, fetch, push, pull, commit |\n | Rust | cargo build/test/clippy/check/nextest |\n | JavaScript | npm/pnpm/yarn/bun install/test/audit/outdated, tsc, eslint, vitest |\n | Python | pytest, ruff, mypy, pip |\n | Go | go test (incl. `-json` stream), go build, go vet, golangci-lint |\n | Cloud | aws, terraform plan/apply/init, gcloud |\n | Containers | docker/podman ps/images/build, kubectl get/describe/logs/apply |\n | JVM | gradle build/test, maven |\n | System | grep/rg, tree, find/fd, ls, curl/wget |\n | GitHub | gh pr/issue/run (JSON + table) |\n\n Unknown commands fall through to the generic compression pipeline — no output is ever left uncompressed.\n\n2. **Structural summaries** — code files compressed to imports + function signatures + call graph (~70% reduction). The model sees the architecture, not implementation noise.\n3. **Dedup cache** — SHA-256 content hash, persistent across sessions. Second read = 13-token reference.\n4. **JSON pipeline** — strip nulls → project out debug fields → flatten → collapse arrays → TOON encoding (lossless compact format)\n5. **Safe mode** — stack traces, secrets, migrations detected by entropy analysis and routed through with 0% compression\n\nFor the full technical details, see [docs/](https://github.com/ojuschugh1/sqz/blob/main/docs/).\n\n## Configuration\n\n```toml\n# ~/.sqz/presets/default.toml\n[preset]\nname = \"default\"\nversion = \"1.0\"\n\n[compression.condense]\nenabled = true\nmax_repeated_lines = 3\n\n[compression.strip_nulls]\nenabled = true\n\n[budget]\nwarning_threshold = 0.70\ndefault_window_size = 200000\n```\n\n### Per-project database\n\nEverything sqz persists (stats, dedup cache, sessions) lives in one SQLite\nfile, `~/.sqz/sessions.db` by default. Set `SQZ_DB_PATH` to keep it\nper-project instead:\n\n```sh\n# e.g. in the project's .envrc (direnv)\nexport SQZ_DB_PATH=\"$PWD/.sqz/sessions.db\"\n```\n\nEvery surface honors it — the shell hook, `sqz stats`/`gain`/`expand`, and\nthe MCP server (set it under `env` in your MCP config). Prefer absolute\npaths: relative values resolve against whatever directory the process runs\nfrom. The parent directory is created if missing (mode 0700, like `~/.sqz`).\n\n## Privacy\n\n- Zero telemetry — no data transmitted, no crash reports\n- Fully offline — works in air-gapped environments\n- All processing local\n\n## Development\n\n```sh\ngit clone https://github.com/ojuschugh1/sqz.git\ncd sqz\ncargo test --workspace\ncargo build --release\n```\n\n## License\n\n[Elastic License 2.0](https://github.com/ojuschugh1/sqz/blob/main/LICENSE) (ELv2) — use, fork, modify freely. Two restrictions: no competing hosted service, no removing license notices.\n\n## Links\n\n- [White Paper: Pre-Injection Context Compression](https://github.com/ojuschugh1/sqz/blob/main/docs/whitepaper.md)\n- [Token Savings Benchmark](https://github.com/ojuschugh1/sqz/blob/main/docs/benchmark.md)\n- [Discord](https://discord.gg/j8EEyH5dSB)\n- [Changelog](https://github.com/ojuschugh1/sqz/blob/main/CHANGELOG.md)\n- MCP Registry name: `mcp-name: io.github.ojuschugh1/sqz`\n\n## Star History\n\n\n

\n\n\n## Contributors\n\nThanks to everyone who has contributed code, fixes, and ideas to sqz:\n\n\n $\"sqz$ \n\n\nAnd to everyone who filed the detailed bug reports behind our fixes. Precise repros make this project better with every release. Want to join them? PRs are reviewed fast: see the [open issues](https://github.com/ojuschugh1/sqz/issues) to get started.\n\n_{Contributor grid made with [contrib.rocks](https://contrib.rocks).}\n","changelogMarkdown","videoKey","requiredSecretsJson","publishedAt","priceCents","trialDays",0,"iconType","image","upstream",{"_109":110,"_111":112,"_113":114,"_115":116,"_117":64,"_118":119},"requiredSecrets",[],"tags",[94,95,96,97,98,99,100,101,102,103,104,105,106,107,108],"assets",[],"stats",{"_92":79,"_93":79},"ratingAvg","reviewCount","agentic-ai","ai","cli","context","cost-optimization","developer-tools","extensions","javascript","llms","open-source","python","rust","token","token-optimization","tokens","topics",[94,95,125,96,97,98,99,100,101,102,103,104,105,106,107,108],"license","NOASSERTION","language","Rust","registryName","io.github.ojuschugh1/sqz","homepage","packages",[120],{"_121":122,"_123":124,"_69":70},"registryType","cargo","identifier","sqz-mcp","api","ENV",{"_128":129,"_130":131,"_132":133},"ENVIRONMENT","production","TURNSTILE_SITE_KEY","0x4AAAAAAB8-rEI2gNDf_QhU","SENTRY_DSN","https://45fd1229f78089d5c3ee78be128030e1@o4510583261757440.ingest.de.sentry.io/4510584253448272"]