context-window plugins

by Mert Koseoğlu

MCP server that saves 98% of your context window with session continuity. Sandboxed code execution in 11 languages, FTS5 knowledge base with BM25 ranking, and automatic state restore across compactions.

by ooples

Context-window optimization: caching, compression, and smart file tooling that cut token usage 60-90%.

by thousandflowers

Measures what share of your agent's loaded context is ever actually invoked - skills, MCP servers, subagents, hooks, always-loaded prose - using evidence from your own session transcripts. Then prunes the dead weight, reversibly.

by Alberto Leal

Configure the Claude Code status line with VCS-aware scripts showing git branch, jj change ID, bookmarks, context usage, and costs. Use when setting up a statusline, customizing the status bar, or adding VCS info to the status line.

by Melodic Software

Measure a Claude Code session's fixed startup context payload per item, on the consumer's machine at a pinned, version-stamped binary, including per-tool attribution of the built-in tool pools that /context reports only as lump sums, derived live by A/B bare-name-deny differencing with enforced comparability rules (skill-listing signature, one mode, one binary), an SDK-primary exact meter degrading to a version-aware headless /context parser and then to an honest structured error (never a wrong number), and a per-project measure-toggle-remeasure ledger under the plugin data directory recording every lever's real before/after delta. Report-only: prints exact config, applies nothing.

by Melodic Software

Per-session context-window observability plus the first shipped consumer: a statusline wrapper tees each session's context_window fields to a per-session snapshot file, a zone resolver classifies usage into smart/acceptable/dumb bands (percentage bands plus window-class token bands, conservative-min combination, zones.json SSOT with shipped defaults), a reader contract fixes how consuming sessions interpret the snapshots, and zone-crossing hooks report once per transition into a worse zone across two channels: the continuation menu to the operator, who owns that choice, and to the model only the zone determination plus the counter-steer that a zone word is not a decay signal (advisory by default; an optional blocking mode gates new mutating work on a fresh dumb-zone snapshot with handoff-writing exempt), with a PostCompact hook persisting an evidence-degraded marker.

by Scott Converse

Context window optimization for Claude Code / Cowork. Sandboxes tool output, compresses what returns with a self-learning 3-stage pipeline, and routes tools automatically — reducing context consumption by 30–60% in typical sessions, more in research-heavy ones.

by cjthompson

Reduced-toolset sub-agent profiles (read-only, lean-executor, standard-executor, main, full-executor) that lower System-tools token overhead vs. spawning the default agent. Pair with the project-tasks skill, which dispatches scouts, executors, and verifiers to these agents.

by Tarcísio Jr

Roteia leituras grandes do Claude Code para um modelo Ollama local. Os hooks bloqueiam Read/cat/sed acima do limiar, inclusive fatiadas, e apontam para o bulk-read, que devolve bullets ancorados em linha.

by Zvi Rahamim

Measure what an AI coding session actually costs and get advice derived from your own transcripts — where the tokens went, which were re-read on every turn, and which fixes are worth making on this machine. Tells one person to install a compressor and the next that a compressor would be worthless, because it reads their numbers first.