Skip to main content
Glama
lcc (Local Context Compiler)

lcc (Local Context Compiler)

lcc — Local Context Compiler

Compact coding-agent context without summarizing it away. lcc cuts what the model does not need, keeps everything else byte for byte, and writes a receipt.

PyPI Python CI MIT DOI GitHub stars

/compact vs lcc

Claude Code's /compact replaces your session with a summary written by a model. Summaries drop the exact error, path or line the agent needs two turns later. lcc does the opposite: old tool outputs leave the context but stay one recall away, and every user and assistant message is kept as it was.

One real Claude Code session (193 messages, 75 tool calls)

tokens

user and assistant messages

removed tool outputs

before

57,083

43, as written

—

/compact

one summary

rewritten into the summary

gone

lcc (rules, stashed)

28,807 (−49.5%)

43 / 43 byte for byte

33, each one recall away

Measured with lcc measure-session (PyPI 1.2.0, offline, exact o200k_base token count; re-run 2026-10-23). On main, whose stash note now states lines and size, the same session gives 28,983 (−49.2%). That session is one of the best cases. Across all 299 of my long sessions (100+ messages each, 36.3 M tokens) the median cut was 21.8% (middle half 15% to 31%, best 65%), with 0 messages lost (python benchmarks/research/long_sessions.py, 2026-10-23, main). These are all my sessions: run it on yours and post the result.

In a pre-registered test on 172 real requests (evidence), an agent answering from the stashed context plus recall did as well as with the full context on file recall (+0.006, 95% CI −0.023 to +0.034) with 36.6% fewer tokens. It called recall for 90% of the questions about removed content. A /compact-style summary of the same sessions got 73% of exact values right (our reimplementation of /compact, written by Opus); full and verbatim context got 98%.

Related MCP server: longhaul

Install

pip install "local-context-compiler[tiktoken]"   # the CLI is `lcc`
lcc fetch-tokenizer                               # once: exact token counts
lcc measure-session "$(ls -t ~/.claude/projects/*/*.jsonl | head -1)"   # your latest session

The PyPI package is local-context-compiler. pip install lcc installs an unrelated astronomy package that happens to own that name.

fetch-tokenizer downloads tiktoken's encoding files once; it is the only lcc command that goes online, and without it counts are labelled approximate. The last line reads your most recent Claude Code session, runs the rules modes offline in seconds and prints counts only. Nothing is changed and nothing leaves the machine.

Claude Code plugin (replaces the /compact summary). Function hooks are early access, so this needs Claude Code 2.1.274+ and the flag below:

export CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1
claude plugin marketplace add lucasmartins-ai/lcc
claude plugin install lcc@lcc --config provider=rules --config rulesMode=stashed

Allow mcp__plugin_lcc_lcc__recall so the agent can get removed outputs back. When a pass fails or removes too little, the plugin hands over to the built-in summary and says so in a toast. Details: Claude Code guide.

MCP (Claude Code, Cursor, Claude Desktop, any stdio client):

{"mcpServers": {"lcc": {"command": "lcc", "args": ["mcp"]}}}

Tools: compact, compact_transcript, recall, inspect, prepare, explain, intake (MCP guide).

CLI, on a session file or any document:

lcc compact ~/.claude/projects/<project>/<session>.jsonl -q "finish the parser fix" \
  --mode tool-calls --provider rules --rules-mode stashed -o compacted.json -r report.json
lcc recall <id>                                   # a removed output, hash-checked

lcc compact notes.md -q "what is wrong with the booking flow?" \
  --provider mechanical -o compacted.md -r report.json
lcc explain report.json --source notes.md         # why each block was kept or dropped

Send us your numbers

lcc measure-session ~/.claude/projects/<project>/<session>.jsonl --markdown
# Codex CLI rollouts are auto-detected from their session_meta record:
lcc measure-session /path/to/rollout.jsonl --markdown
# or a whole project directory: per-mode median and p90 reduction over its sessions
lcc measure-session ~/.claude/projects/<project>/ --markdown

It runs every mode on one of your sessions, offline, and prints counts only: messages, tool calls, tokens before and after, messages kept, time. No text, path or tool output. For exact counts, install the tiktoken extra and cache its tokenizer once (lcc itself never downloads): python -c "import tiktoken; tiktoken.get_encoding('o200k_base')". Otherwise the report says approximate and why. Paste it in Discussions → Results. Add --input-map to see where the session's billed input goes instead: cache reads and writes, and the share taken by tool results, tool-call inputs, injected instructions and the fixed system prompt and tool schemas (counts only, from the usage records in the file). Already compacting with lcc? lcc share-stats turns your local reports into the same kind of counts-only summary (SHARE_STATS.md), or use the results issue form. Every number from someone other than me makes the claims above less mine and more true.

Why lcc

  • Verbatim. Kept spans are byte-identical to the source. Nothing is summarized or rewritten.

  • Recoverable. stashed moves removed outputs to a local content-addressed store (~/.cache/lcc/stash); recall re-hashes on read. The placeholder, [removed: 120 lines, 4.3 KB · lcc recall <id>], states lines and size, not tokens, so it is identical on every machine.

  • Auditable. Every pass writes a JSON report; lcc explain gives the reason for every drop.

  • Local. The default judges run on your CPU in plain Python. No API key, no network.

Providers

Provider

Runs

When it is used

Report says

specialist-v3

local, CPU

default (auto) whenever transcript blocks are present: the Claude Code hook, compact_transcript, --mode tool-calls

semantic_guarantee: none (a rank, not a judgment)

mechanical

local, CPU

--provider mechanical; plain text under auto (a configured TypeSafe key does not change that unless LCC_AUTO_REMOTE=1)

semantic_guarantee: none

rules (lossless, recoverable, stashed)

local, no model

provider: rules for transcripts

lossless: only provable duplicates removed; recoverable: also outputs that can be re-read; stashed (recommended): removed outputs stay byte-recoverable through recall

jev

remote (TypeSafe API)

opt-in: --provider jev / provider: jev, needs TYPESAFE_API_KEY

semantic judgment or explicit degradation

laya

local model

deprecated; runs only when named

named model and fallback state

specialist-v3 details and limits: SPECIALIST.md. Rules modes: TOOL_CALLS.md. Jev is an optional path you name explicitly (--provider jev); it sends the objective and context blocks to the TypeSafe API, and LCC_DISABLE_NETWORK=1 disables it. auto never picks it on its own: having a key is not consent to send context off-machine, so plain text under auto stays on mechanical unless you set LCC_AUTO_REMOTE=1. compact writes a relevance-compaction-1.2 report; check provider_used and degraded before trusting the output.

Benchmarks

Real sessions (held-out, INFERIDO). In plain words: when a ranker has to drop part of a session, how often does it keep what the agent actually used next? 1.0 would be always. A one-shot final test on 193 items from 36 of the maintainer's sessions that started after the training cutoff. Retention AUC over drop budgets of 0–60% of candidate tokens:

ranker

recall AUC

all-kept AUC

specialist-v3

0.899

0.434

mechanical

0.829

0.238

recency

0.818

0.306

bm25

0.803

0.270

random

0.689

0.138

The labels are next-action evidence from one maintainer's Claude Code sessions, which favours v3's training target. The sessions are private; only aggregates are published (SPECIALIST.md).

Synthetic, reproducible (python benchmarks/run_judge_bench.py, about 20 s, no network). Nine hand-written edge cases: supersession, contradiction, config change, exception rule, pt/en mixed, prompt injection in a tool output, large git diff, old root cause, re-read.

raw ranking

specialist-v3

mechanical

bm25

recency

cases with every must-keep block kept at a 30 / 50 / 70% cut (of 9)

8 / 7 / 7

9 / 9 / 9

9 / 9 / 7

8 / 6 / 3

full pass, default settings

must-keep kept

302 blocks

1,138 blocks

peak RSS

specialist-v3, tool-call mode

9/9

181 ms

3.4 s

71 MB

rules lossless

9/9

2 ms

16 ms

55 MB

These cases are lexically easy, so the lexical baselines do well; v3 misses the Portuguese case. Timings are from macOS arm64 under heavy machine load. Method, every arm and the limits: BENCHMARKS.md. Rules modes on real sessions: REAL_SESSION_BENCHMARKS.md.

Reduction depends on how much of a context is noise. Below a 25% gain, the report marks the pass as not worth it, and the Claude Code plugin keeps the built-in summary.

What it does not do

  • It does not check that a model will answer correctly. A structural PASS is not an answer quality result.

  • No RAG, embeddings or vector store.

  • Protection rules and bounded restoration reduce, but do not rule out, lost evidence; see the safety model.

  • specialist-v3 was trained on Claude Code transcripts. Other agents and plain documents are out of distribution; plain text uses mechanical unless you opt in to Jev.

Docs

Topic

Link

What lcc is, the glossary, product vs research

CONCEPT.md

Five-minute tour of the CLI

QUICKSTART.md

Claude Code plugin

CLAUDE_CODE.md

MCP server

MCP.md

Tool-call compaction and rules modes

TOOL_CALLS.md

Aider chat history as input

AIDER.md

The local ranker

SPECIALIST.md

Benchmarks

BENCHMARKS.md, REAL_SESSION_BENCHMARKS.md, benchmarks/

Python API (lcc.msi.compile)

msi-api.md

Architecture and decisions

architecture.md, ADR 0010, ADR index

Opt-in and deprecated providers

JEV.md, LAYA.md

Research archive

paper.md, results-master.md, TRANSCRIPT_AB.md

Privacy

PRIVACY_POLICY.md

Development

Python 3.11+.

python -m pip install -e '.[dev,tiktoken]'
pytest -q && ruff check . && mypy
npm test                                   # Node API and hook tests
python benchmarks/run_judge_bench.py --check --out /tmp/judge-bench

The PyPI package named lcc is unrelated; this one is local-context-compiler. Contributing: CONTRIBUTING.md. Security: SECURITY.md. Changes: CHANGELOG.md.

License

MIT. The optional Laya integration carries Apache-2.0 attribution in NOTICE. Maintained by LookADev.

Available Tools

7 tools
compactB

Drop context blocks irrelevant to an objective. Default provider mechanical is fully offline (no key, no network); laya is offline semantic (needs the [laya] extra, else honest mechanical fallback); jev needs TYPESAFE_API_KEY. Fail-safe: failures keep content.

Inputs: text (required), question (required), provider, threshold.

compact_transcriptA

Tool-call compaction for a session transcript: pairs each tool call with its result, pins the first and newest messages, and drops the pairs a semantic judge says are spent — kept messages stay verbatim, so nothing is summarized. Default 'auto' is the local specialist-v3 ranker (offline, no key); 'jev' is remote and needs TYPESAFE_API_KEY. Every failure keeps the call.

Inputs: messages (required), question (required), ledger, trigger, +20 more.

explainB

Explain a relevance-compaction report (as produced by compact): why every block was kept, trimmed, or dropped. Never re-runs, never touches the network.

Inputs: report (required), only, limit, source_text.

inspectA

Read-only diagnostic inspection of a text: tokens, structure, duplication, cleanup projection, and recommendation. Offline.

Inputs: text (required).

intakeB

Intake triage for raw prompt input: readiness with named reasons, non-blocking risk_flags, clarifying questions, a receipt of every filler removed from dictation, and a cache-stable compiled prompt. Offline.

Inputs: text (required), question, previous_turn.

prepareC

Inspect-first deterministic prompt preparation: runs inspection, then lexical selection + safe optimization only when recommended. Never summarizes or paraphrases. Offline.

Inputs: text (required), question.

recallA

Return a tool output that compact_transcript (rules_mode 'stashed') replaced with '[removed: N lines, S KB · lcc recall ]'. The bytes are re-hashed on read, so the text is exactly what was removed. Offline, local stash only.

Inputs: id (required), file, lines, ledger.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.5/5.0

    Scored across 7 tools

    Disambiguation4/5

    Most tools have distinct scopes: intake handles raw prompts, inspect is read-only diagnostics, prepare does deterministic prompt preparation, compact operates on context blocks, and compact_transcript operates on session transcripts. The boundaries are mostly clear, though compact vs compact_transcript and inspect vs prepare could be confused at first glance, requiring the descriptions to disambiguate.

    Naming Consistency4/5

    All names are lowercase snake_case and are mostly concise verb-style names. The set is mostly consistent, with minor deviations such as compact_transcript being a compound name while most others are single verbs, and intake reading more like a noun than a verb.

    Tool Count5/5

    Seven tools is well-scoped for a local context compiler. Each tool appears to earn its place across intake, inspection, preparation, compaction, explanation, and recall, with no obvious filler tools.

    Completeness4/5

    The surface covers the main lifecycle: intake, inspect, prepare, compact, compact_transcript, explain, and recall. Minor gaps exist, such as no obvious listing or management operations for stashed outputs or generated reports, but the core workflows are covered and agents can work around these gaps.

    Maintenance

    ActivityActive
    ResponsivenessLimited data

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      B
      maintenance
      Provides deterministic context compaction and safety gating for AI agents, preserving original messages verbatim while dropping irrelevant ones and flagging destructive commands.
      2
      MIT
    • A
      license
      Not graded
      quality
      Not graded
      maintenance
      Enables agents to deterministically shrink multi-turn conversation history and prompts by stripping verbose boilerplate and filler tokens while preserving critical tool parameters and named entities. This lets users cut inference costs by roughly 40-60% without losing essential context, running either as a native MCP server over JSON-RPC stdio or as an importable zero-dependency Python module.
      MIT