Show HN: Lcc – compact Claude Code sessions without summarizing them

A tool called lcc, published to PyPI as version 1.2.0 by developer lucasmartins-ai, compacts Claude Code sessions by removing old tool outputs while keeping every user and assistant message byte for byte, rather than replacing the session with a model-written summary as Claude Code's built-in `/compact` does. On one 193-message session with 75 tool calls, lcc cut tokens from 57,083 to 28,807 (−49.5%) while preserving all 43 user and assistant messages and stashing 33 removed tool outputs for retrieval via a `recall` command; across 299 long sessions (100+ messages each, 36.3M tokens) the median cut was 21.8% with 0 messages lost. In a pre-registered test on 172 real requests, an agent answering from the stashed context plus `recall` matched full-context performance on file recall (+0.006, 95% CI −0.023 to +0.034) with 36.6% fewer tokens, calling `recall` for 90% of questions about removed content, while a `/compact`-style summary of the same sessions got 73% of exact values right versus 98% for full verbatim context.

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. 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 send-us-your-numbers and post the result. In a pre-registered test on 172 real requests evidence https://github.com/lucasmartins-ai/lcc/blob/main/docs/EVIDENCE.md , 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%. pip install "local-context-compiler tiktoken " the CLI is lcc lcc measure-session "$ ls -t ~/.claude/projects/ / .jsonl | head -1 " your latest session The second line reads your most recent Claude Code session, runs every mode offline 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 https://github.com/lucasmartins-ai/lcc/blob/main/docs/CLAUDE CODE.md . 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 https://github.com/lucasmartins-ai/lcc/blob/main/docs/MCP.md . 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 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 https://github.com/lucasmartins-ai/lcc/discussions/categories/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 https://github.com/lucasmartins-ai/lcc/blob/main/docs/SHARE STATS.md , or use the results issue form https://github.com/lucasmartins-ai/lcc/issues/new?template=results.yml . Every number from someone other than me makes the claims above less mine and more true. - 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. | 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 https://github.com/lucasmartins-ai/lcc/blob/main/docs/SPECIALIST.md . Rules modes: TOOL CALLS.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/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. 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 https://github.com/lucasmartins-ai/lcc/blob/main/docs/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 https://github.com/lucasmartins-ai/lcc/blob/main/docs/BENCHMARKS.md . Rules modes on real sessions: REAL SESSION BENCHMARKS.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/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. - 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 https://github.com/lucasmartins-ai/lcc/blob/main/docs/adr/0014-minimum-sufficient-context.md . - 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. | Topic | Link | |---|---| | Five-minute tour of the CLI | QUICKSTART.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/QUICKSTART.md | | Claude Code plugin | CLAUDE CODE.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/CLAUDE CODE.md | | MCP server | MCP.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/MCP.md | | Tool-call compaction and rules modes | TOOL CALLS.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/TOOL CALLS.md | | Aider chat history as input | AIDER.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/AIDER.md | | The local ranker | SPECIALIST.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/SPECIALIST.md | | Benchmarks | BENCHMARKS.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/BENCHMARKS.md , REAL SESSION BENCHMARKS.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/REAL SESSION BENCHMARKS.md , benchmarks/ https://github.com/lucasmartins-ai/lcc/blob/main/benchmarks/README.md | | Python API lcc.msi.compile | msi-api.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/lcc/msi-api.md | | Architecture and decisions | architecture.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/architecture.md , ADR 0010 https://github.com/lucasmartins-ai/lcc/blob/main/docs/adr/0010-deterministic-first-preparation-model-assistance.md , ADR index https://github.com/lucasmartins-ai/lcc/blob/main/docs/adr/README.md | | Opt-in and deprecated providers | JEV.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/JEV.md , LAYA.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/LAYA.md | | Research archive | paper.md https://github.com/lucasmartins-ai/lcc/blob/main/research/paper.md , results-master.md https://github.com/lucasmartins-ai/lcc/blob/main/research/results-master.md , TRANSCRIPT AB.md https://github.com/lucasmartins-ai/lcc/blob/main/benchmarks/research/TRANSCRIPT AB.md | | Privacy | PRIVACY POLICY.md https://github.com/lucasmartins-ai/lcc/blob/main/docs/PRIVACY POLICY.md | 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 https://github.com/lucasmartins-ai/lcc/blob/main/CONTRIBUTING.md . Security: SECURITY.md https://github.com/lucasmartins-ai/lcc/blob/main/SECURITY.md . Changes: CHANGELOG.md https://github.com/lucasmartins-ai/lcc/blob/main/CHANGELOG.md . MIT https://github.com/lucasmartins-ai/lcc/blob/main/LICENSE . The optional Laya integration carries Apache-2.0 attribution in NOTICE https://github.com/lucasmartins-ai/lcc/blob/main/NOTICE . Maintained by LookADev https://lookadev.com .