简介

OpenLore

<p align="center"> <strong>Deterministic, local-first memory and guardrails for AI coding agents — with no LLM in the hot path.</strong><br> One call tells your agent the code a task touches; one gate tells it what's unsafe to change.<br> Grounded in static analysis. No API key. Same answer every time. </p>

<p align="center"> <a href="https://www.npmjs.com/package/openlore"><img src="https://img.shields.io/npm/v/openlore?color=2563eb&label=npm&logo=npm&logoColor=white" alt="npm version"></a> <a href="https://github.com/clay-good/OpenLore/actions/workflows/ci.yml"><img src="https://github.com/clay-good/OpenLore/actions/workflows/ci.yml/badge.svg" alt="CI status"></a> <a href="LICENSE"><img src="https://img.shields.io/npm/l/openlore?color=22c55e" alt="MIT License"></a> <img src="https://img.shields.io/node/v/openlore?color=339933&logo=node.js&logoColor=white" alt="Node >=22.19"> <br> <img src="https://img.shields.io/badge/MCP-ready-7c3aed?logo=anthropic&logoColor=white" alt="MCP ready"> <img src="https://img.shields.io/badge/languages-21%20%2B%2012%20IaC-f97316" alt="21 languages + 12 IaC ecosystems"> <img src="https://img.shields.io/badge/tests-5500%2B-success" alt="5500+ tests"> <img src="https://img.shields.io/badge/API_key-not_required-0ea5e9" alt="No API key required"> <a href="https://github.com/clay-good/OpenLore/stargazers"><img src="https://img.shields.io/github/stars/clay-good/OpenLore?style=social" alt="GitHub stars"></a> </p>

<p align="center"> <img src="docs/openlore-demo.gif" alt="An unedited terminal recording of the published openlore v2.1.6 on a fresh clone of ripgrep: openlore install wires the agent and indexes 235 files live in 14 seconds with no API key, openlore orient returns the functions a task touches, openlore review reports a changed signature that left 39 callers stale, and openlore prove --estimate projects the payoff on that repo" width="100%"> </p>

<p align="center"><em>A real, unedited recording — the published <code>openlore</code> on a fresh clone of <a href="https://github.com/BurntSushi/ripgrep">ripgrep</a>. <strong>install</strong> wires your agent and indexes the repo live — 235 files, 2,978 functions, 4,329 call edges in <strong>14 seconds</strong>, no API key → <strong>orient</strong> returns the code a task touches → <strong>review</strong> catches a signature change that left <strong>39 callers</strong> stale → <strong>prove</strong> projects the payoff. Re-record it yourself: <a href="docs/openlore-demo.tape"><code>docs/openlore-demo.tape</code></a>.</em></p>

<p align="center"> <strong><a href="#install-in-one-command">Install</a> · <a href="#what-you-get">What you get</a> · <a href="#value-scorecard--does-it-pay-for-itself">Benchmarks</a> · <a href="#governance">Governance</a> · <a href="#how-it-works">How it works</a> · <a href="#openlore-vs-alternatives">vs. Alternatives</a> · <a href="#documentation">Docs</a></strong> </p>

---

AI coding agents are powerful but amnesiac and ungoverned: every task restarts by re-reading the same files, long sessions drift onto stale assumptions, and nothing warns the agent when a change is about to break a contract or cross a boundary.

OpenLore fixes both halves. It runs a one-time static analysis of your repo and keeps a live knowledge graph — call structure, types, tests, decisions, IaC, spec drift. Your agent queries it to start every task already oriented and to certify a change before it lands. It's deterministic and local-first — no LLM in the hot path — so the same question always returns the same grounded answer, and the agent is told when a fact goes stale instead of served a confident guess.

Core features

Everything is deterministic and local; only the two entries marked "API key" ever talk to a model.

Analyze (no key) — Full call graph in SQLite, community detection, McCabe complexity, extracted DB schemas / HTTP routes / UI components / middleware / env vars. Outputs a ~600-token CODEBASE.md digest. The file watcher updates the graph incrementally on every save and converges to what analyze --force would produce; when a change exceeds the per-save budget, the un-recomputed files are marked explicitly stale, never silently divergent.

Drift (no key) — Compares git changes against spec mappings in milliseconds (Gap / Uncovered / Stale / ADR-gap). Installs as a pre-commit hook. → docs/drift-detection.md

Test-impact selection (no key) — select_tests walks the call graph backward from a change to every test that reaches it, with paths. An honest over-approximate prioritizer, not a replacement for the full suite. → docs/test-impact-selection.md

<details> <summary><strong>All the other tools</strong> — certificates, dead-code, invariants, coverage gaps, clones, error & env impact, coupling…</summary>

• find_dead_code (no key) — cross-language mark-and-sweep, "what dies if I delete X?" Confidence-tagged candidates, never deletion authority. → docs/reachability-dead-code.md
• change_impact_certificate (no key) — certifies whether a diff newly opens a path into a declared surface (differential reachability), plus blast radius and tests. CLI: openlore impact-certificate.
• certify_public_surface (no key, opt-in) — breaking-change verdict per export, consumers named. Renamed exports detected via symbol-identity continuity. CLI: openlore certify-public-surface.
• check_architecture (no key) — enforces author-declared layer, boundary, dependency, cycle, reachability, orphan, and instability rules. → docs/architecture-invariants.md
• verify_claim (no key, opt-in) — confirmed / refuted / unverifiable with a citation receipt, never an LLM guess.
• openlore enforce (no key, advisory) — the unified gate over all governance findings; one enforcement.policy maps each finding → blocking / frozen / advisory / off. → docs/configuration.md
• Decisions on the graph (API key for consolidation) — record_decision before writing code; a pre-commit hook gates until reviewed. Decisions become decision:: nodes joined to the files they govern, so analyze_impact returns them as neighbors.
• Epistemic Lease (no key) — models drift as a navigation phenomenon; every MCP response carries a brief, factual freshness note once context ages. orient() resets it.
• structural_diff (no key) — the structural complement to git diff: functions/edges added/removed, signature changes, and callers now stale. → docs/structural-diff.md
• get_change_coupling (no key) — co-change coupling and churn from git. Advisory, correlation not causation. → docs/change-coupling.md
• report_coverage_gaps (no key, opt-in) — which load-bearing code has no test reaching it, ranked by significance. Never claims a symbol is "tested." → docs/coverage-gaps.md
• get_style_fingerprint — a descriptive idiom profile so an agent matches the house style; a counter below the evidence floor reports null, never a guess.
• find_clones — the edit-time "does a near-duplicate already exist?" query (a symbol or raw snippet), ranked exact > structural > near.
• analyze_error_propagation — exceptions that escape vs. those caught (TS/JS/Python/Java/C#), or returned errors and panic/recover flow in Go; a sound lower bound.
• analyze_env_impact — "what breaks if I remove this env var?": read sites, upstream callers, tests, per-site required.
• briefing_since — the catch-up lens: changed symbols since a base ref, ranked into a fixed tier order.
• plan_parallel_work / map_in_flight_conflicts — a hazard-typed conflict graph over a task list, or over every in-flight branch/PR/agent-task (opt-in coordination preset).
• Share the index — the graph is a function of committed source, so a team analyzes once: openlore export bundle → a portable .olbundle, openlore import bootstraps in seconds (validate-or-rebuild). → docs/shareable-bundle.md
• Preflight — a CI staleness gate; any PR editing indexed files fails until the graph is refreshed. → docs/preflight.md

</details>

---