Lifeline v0.3.0

CLI reference

Every Lifeline command, generated from the CLI itself: init, context, resume, checkpoint, log, recall, policy, audit, guard, lines, verify, push/pull/clone.

Generated from the parser by python scripts/gen_cli_doc.py — don't edit by hand (a test fails when this file drifts).

Globals, before the command: --db <path> (default .lifeline/ledger.db, or LIFELINE_DB) · --line <name> (a named line: .lifeline/<name>.db + LIFELINE.<name>.md) · --store sqlite|supabase. Identity on a write: --author / --agent / --provider / --model.

Start

Command What it does
lifeline init [--out OUT] [--hooks] [--agent AGENT] [--no-mcp] [--no-policy] one-command setup: the line, .mcp.json, the write policy (and --hooks); then the three lines that bootstrap it
lifeline install-hooks [--local] [--no-guard] [--root ROOT] {claude-code,git} install the auto-connect hook in a client (SessionStart: startup, resume and after each compaction)
lifeline install-rules [--clients CLIENTS] [--root ROOT] write the Lifeline section into each client's rule file (CLAUDE.md, AGENTS.md, GEMINI.md, .cursor/rules, copilot-instructions)
lifeline migrate --from SRC [--strict] ingest an existing markdown into the store

Every day

Command What it does
lifeline context [--budget BUDGET] [--budget-tokens BUDGET_TOKENS] [--tokenizer TOKENIZER] [--format {generic,anthropic,openai,gemini}] [--query QUERY] [--thread THREAD] [--mark-session] print the assembled context
lifeline resume [--thread THREAD] [--json] where work stopped: the live checkpoint of each thread
lifeline checkpoint [--thread THREAD] [--goal ITEM] [--done ITEM] [--next ITEM] [--blocked ITEM] [--tried ITEM] [--summary SUMMARY] [--author AUTHOR] [--agent AGENT] [--provider PROVIDER] [--model MODEL] [--parents PARENTS] [--out OUT] record where work stopped on a thread (goal/done/next/blocked/tried) — the next AI resumes from it
lifeline log --kind KIND --summary SUMMARY [--body BODY] [--author AUTHOR] [--agent AGENT] [--provider PROVIDER] [--model MODEL] [--parents PARENTS] [--thread THREAD] [--out OUT] human: append directly to the line (you're the approver)
lifeline recall [-k K] query the entries most relevant to a query (anchored; reverted marked)

Curate

Command What it does
lifeline propose --kind KIND --summary SUMMARY [--body BODY] [--author AUTHOR] [--agent AGENT] [--provider PROVIDER] [--model MODEL] [--parents PARENTS] [--thread THREAD] propose an entry (HITL) — stays pending until approved
lifeline review list pending proposals (HITL curation)
lifeline approve pids [pids ...] seal pending proposal(s) into the line
lifeline reject pids [pids ...] discard pending proposal(s)
lifeline policy [--init] [--agent AGENT] show the project's write policy (.lifeline/policy.json)
lifeline audit [--since SINCE] [--status {auto,approved,rejected,pending,duplicate,all}] the write trail: policy-direct writes (auto), approved, rejected, pending
lifeline guard (--session \| --git \| --since SINCE) [--require {any,checkpoint}] discipline check: did this line get a checkpoint/entry? --session (Stop hook, blocks once) · --git (pre-push) · --since X
lifeline pin [--body BODY] [--author AUTHOR] [--out OUT] id pin a decision as FOUNDATION of the context (always shown; an anchored note, never a mutation)
lifeline unpin [--body BODY] [--author AUTHOR] [--out OUT] id unpin a decision (a correction superseding the pin note)
lifeline exam [--json] [--budget BUDGET] Context Health: score 0-100 of how ready this line leaves a fresh AI (what/why/decided/next) + concrete gaps
lifeline capture [--last LAST] [--author AUTHOR] draft proposals from recent git commit messages (zero-LLM; the commit body is the why — no body, no draft)

Lines & mesh

Command What it does
lifeline lines [--mesh] list the project's lines (.lifeline/*.db)
lifeline promote --from SRC_LINE --to DST_LINE [--id IDS] [--kind PKIND] copy entries from one line into another as roots (idempotent — re-promoting dedups by content)
lifeline where id which lines of the project hold an id (mesh lookup)
lifeline relate --from FROM_ID --to TO_ID [--rel {depends_on,implements,tests_covers,supersedes}] [--author AUTHOR] [--out OUT] record a typed relation between two entities (depends_on / implements / tests_covers / supersedes)
lifeline graph (--depends-on ID \| --dependents ID \| --of ID) query the typed entity graph (structured answer)

Maintain

Command What it does
lifeline verify [--mesh] [--no-mesh] check the ledger's integrity (parents in sibling lines count as external anchors, not omissions)
lifeline rebuild [--out OUT] regenerate the view from the store
lifeline reindex rebuild the recall index (FTS5) from the ledger
lifeline rehash [--out OUT] recompute every id under the current canonical scheme (one-shot migration after a hash-scheme upgrade)
lifeline schema print the cloud SQL schema (Supabase) — paste it into the SQL Editor

Sync & cloud

Command What it does
lifeline push [--out OUT] git sync: rebuild + commit + push the view
lifeline pull [--out OUT] git sync: pull + rebuild the .db from the merged view
lifeline clone url dest git clone and rebuild the .db
Lifeline — MIT licensed · Edit on GitHub ← Back to the lifeline