or ) or a deterministic stacked pipeline that executes multiple engines in order.

. Engines expose a shared contract:

: stable engine id such as or
• : legacy execution path used by stacked pipelines
• : primary execution path returning text + stats
• : returns the JSON-Schema-like shape of valid config
• : returns

(or for advanced cases), which calls and before accepting. Use to remove an engine at runtime.

registers the built-in engines before compression runs. This lets preview, runtime compression, stacked mode, tests, and future engines use the same execution path.

and MCP-SERVER.md. It reuses Caveman rules but operates on tool metadata, not request payloads.

• , , and modes

.

fewer output tokens, average output savings in benchmarks with a range, and a input-compression tool. OmniRoute uses the Caveman input-side number when documenting stacked prompt/context savings; Caveman output mode remains a separate response-behavior feature.

, , , Vitest/Jest/Pytest, Cargo/Go tests, TypeScript/Vite/Webpack builds, ESLint, npm audit/installs, Docker logs, shell /, stack traces, and generic logs
• and global filters in

.

.

savings for command-output compression. Its README example shows a 30-minute Claude Code session going from tokens to , or saved.

rtk -> caveman

in compression settings or through compression combos.

combined = 1 - (1 - RTK savings) * (1 - Caveman input savings)
average  = 1 - (1 - 0.80) * (1 - 0.46) = 89.2%
range    = 1 - (1 - 0.60..0.90) * (1 - 0.46) = 78.4-94.6%

tool results, not on prompts or context. It targets the verbose accessibility-tree and browser snapshot payloads returned by tools like Playwright, computer-use, and browser-automation MCP servers.

Noise stripping — removes empty generic/text entries (, )
1. Sibling collapse — when ≥ (default 30) consecutive lines are structural repeats, collapses them into the first (default 10) lines + a count summary + the last (default 5) lines
2. Ref preservation — anchors required by Playwright/computer-use are never touched
3. Hard truncation — if the text after collapse still exceeds (default 50,000), truncates with a navigation hint so the agent can continue working

open-sse/services/compression/engines/mcpAccessibility/
  index.ts            ← smartFilterText() entry point
  collapseRepeated.ts ← sibling-collapse algorithm
  constants.ts        ← DEFAULT_MCP_ACCESSIBILITY_CONFIG

in global settings (migration 056). Default config:

{
  "enabled": true,
  "maxTextChars": 50000,
  "collapseThreshold": 30,
  "collapseKeepHead": 10,
  "collapseKeepTail": 5,
  "minLengthToProcess": 2000
}

is and whose length exceeds . It does not affect prompt compression or request payloads.

---

: stores mode, pipeline, RTK config, language config, and default marker
• : maps a compression combo to a routing combo
• and

.

	config)

node --import tsx/esm --test tests/unit/compression/rtk-*.test.ts tests/unit/compression/pipeline-integration.test.ts tests/unit/compression/context-compression-api.test.ts
node --import tsx/esm --test tests/unit/compression/*.test.ts tests/golden-set/*.test.ts tests/integration/compression-pipeline.test.ts tests/unit/api/compression/compression-api.test.ts
node --import tsx/esm --test tests/unit/compression/mcpAccessibility*.test.ts
npm run typecheck:core