📖 Lore

Shared, full-fidelity, session history for every AI agent you use.

Any Agent. Any Session. Any Turn. Anytime.

The session you ran in Claude Code this morning, Openclaw can pull up this afternoon. The debugging thread Cursor worked through last week, Hermes can read like it was sitting right there. Lore gives all of your coding agents one shared memory: across every tool, across every session, and it never expires.

That is the part most "agent memory" misses. Other tools remember things for one agent, inside one app. Lore makes every session any of your agents has ever had searchable by any other agent, anytime. What Codex figured out, Claude Code knows. What you solved on Tuesday is one search away on Friday, in whatever agent you happen to be in.

And yes, it survives compaction. When a context window fills up and compacts, the work isn't gone, it's sitting right there on disk. Lore just makes it reachable again, fully searchable, right where your agents work. Local, private, and yours forever.

https://github.com/user-attachments/assets/b5c0f077-47da-4502-bf78-2ce08abf034f

Inspired by RLM

v0.2.0 release

This release turns Lore from a search box over old transcripts into a small workflow substrate for agents that have to continue real work:

• Schema 5 write support. Current installs can read and write the live store again. lore status --json reports both schemaVersion and supportedSchemaVersion, so stale global installs are obvious before sync, index, or push writes fail.
• Freshness checks. lore status tells an agent whether a source/project window is ready, missing, stale, or possibly unsynced before it writes a brief or handoff from incomplete memory.
• Live detected-source sync. lore sync <source> indexes a detected active transcript tree incrementally. This covers codex (~/.codex/sessions, with archived sessions as a compatibility fallback) and claude-code (~/.claude/projects) without forcing agents to remember raw paths.
• Workflow skills. The npm package now ships lore-recall, lore-brief, lore-handoff, and lore-dev-verification alongside the base lore skill. They include references, examples, eval specs, validators, and committed test reports.
• Package smoke proof. npm run package:smoke builds, packs, unpacks, and validates the real tarball so README, help text, skill folders, executable mode, and package metadata are checked together.

Already installed? Refresh the global command before testing any schema or sync behavior:

npm install -g @jordanhindo/lore@latest
lore help
lore status --json

If your shell cached the old path, open a new terminal or run hash -r and retry lore help.

✨ Why you'll want it

• 🔀 One memory, every agent. Claude Code, Codex, openclaw, Cursor, and Hermes histories all land in one store (separate from their own /. stores), and any MCP client reads it. A session in one agent is instantly readable by all the others. What Codex figures out, Claude Code knows.
• 🧠 It outlives the context window. Compaction stops meaning amnesia. Last week's reasoning is one search away.
• 🔍 Search that speaks code. getUserById, foo.bar.ts, and trust-metadata stay findable as whole words, not shredded into noise.
• 🔒 Local by default. Everything lives at ~/.lore/lore.db on your machine. Nothing phones home. Ever.

🚀 Quick start (paste this into your agent)

Lore sets itself up. Drop the blurb below into any coding agent. It installs Lore, indexes your own history, and proves search works before it calls itself done. It is server-free by default, with no MCP registration required.

Set up Lore (full-fidelity session memory) for yourself. Install it with
`npm install -g @jordanhindo/lore@latest`, then run `lore setup` to index my
history and self-verify. Prove it by running `lore search` for a word from
a past session and showing me a hit. The MCP server is optional: only
register it (`lore serve`) if you specifically want MCP tool calls instead
of the CLI; if you do, tell me the reload step I have to do.

The deterministic recipe the agent follows lives in AGENT-ONBOARD.md, which keeps MCP registration as an optional step. Prefer the skill instead? npx skills add jordanhindo/lore (below) drops in the self-bootstrapping lore skill, which does all of this the next time you ask it to recall something.

🛠️ Install it yourself

Recommended: install the skill. It sets up the rest.

npx skills add jordanhindo/lore

This drops the bundled lore skill into your agent (~/.claude/skills/). The skill is self-bootstrapping: the next time you ask your agent to "remember" / "recall" something or to "set up lore", it reads its own references/setup/index.md, installs the lore CLI, indexes your history, and proves search works. No MCP server required. One command installs the whole thing.

Or install the CLI directly:

npm install -g @jordanhindo/lore@latest   # puts the `lore` command on your PATH

Requires Node 22+. That's the whole install. lore setup takes it from here. If npm install chokes on better-sqlite3, or lore isn't found afterward, the Troubleshooting section has you covered, Windows included.

Upgrade the global CLI

The global lore command is just an npm binary shim. If you were running an old package, update it before you trust lore sync, lore index, or lore push:

npm install -g @jordanhindo/lore@latest
lore help
lore status --json

If lore status --json shows a store schemaVersion higher than supportedSchemaVersion, your command is still stale. Reinstall the package, open a fresh shell if needed, then rerun lore status --json.

From source

Cloning is for contributors, or for writing a new adapter (which needs the source tree):

git clone https://github.com/jordanhindo/lore && cd lore
npm install
npm run build
npm link        # optional: puts `lore` on your PATH

📥 Backfill your history

lore setup is the fast path. It sniffs out the harnesses on your machine, indexes each one, checks that search works, and prints how to wire Lore into your MCP client.

lore setup

Want finer control? Point lore index at a transcript directory. For Claude Code that's ~/.claude/projects:

lore index ~/.claude/projects              # primary transcripts
lore index ~/.claude/projects --subagents  # include subagent transcripts
lore index ~/.claude/projects --no-redact  # keep credentials verbatim (see Privacy)

Other harnesses come in with --source. Codex, openclaw, Cursor, and Hermes are all built in:

lore index ~/.codex/sessions --source codex
lore index ~/.openclaw/sessions --source openclaw
lore index ~/.cursor --source cursor     # reads Cursor's SQLite state store
lore index ~/.hermes --source hermes     # reads Hermes's SQLite state store

Cursor and Hermes keep their history in a SQLite database rather than JSONL files; point lore index at the directory and the adapter reads the store directly. Current sampled Cursor rows expose toolResults fields but only empty arrays, so the Cursor adapter indexes text only and fabricates no tool calls.

Re-run any of these whenever. Unchanged files get skipped, so repeat runs are cheap.

Some harnesses write transcript trees instead of emitting Lore-compatible hook payloads. For manual active-session catch-up, use the incremental sync command:

lore sync codex
lore sync claude-code
lore sync hermes

codex sync indexes ~/.codex/sessions incrementally, with ~/.codex/archived_sessions only as a compatibility fallback. claude-code sync indexes ~/.claude/projects incrementally, including subagent transcripts. hermes sync indexes ~/.hermes incrementally.

For cron, launchd, Task Scheduler, or any unattended timer, use the bundled lock-protected wrapper instead of raw lore sync so multiple source jobs do not write ~/.lore/lore.db at the same time:

./scripts/lore-sync-once.sh codex
./scripts/lore-sync-once.sh claude-code
./scripts/lore-sync-once.sh hermes

💻 Recall from the CLI, no server required

You don't need the MCP server to use Lore. The lore command opens the SQLite store directly, so search, read-back, navigation, and writing all work server-free anywhere you have a shell. The MCP server below is just one more reader of the same store; nothing here depends on it.

The loop is always the same: drill down, never dump. A session can be millions of tokens, so you never page a whole one into context. Search broadly, take the real ids it hands back (you never invent one), then spend them:

lore search "fts tokenizer" --relevant     # → hits, each leading with a message + session id
lore get <message-id> --full               # read that one message in full
lore context <message-id>                  # 5 before / 5 after, anchor flagged
lore session <session-id> --around <message-id>   # jump to that spot, read forward

Other commands round out the loop: lore status (freshness/health), lore sessions (recent conversations), lore timeline (activity by day/hour), and lore push (add a live session from a JSON batch on stdin). Add --json to any of them for a machine-readable envelope. Every search filter the MCP tools accept works here too (--project, --source, --session, --branch, --agent, --skill, --tool, --role, --model, --since, --until, --limit).

Before asking for a current brief or handoff, check freshness:

lore status --json --source codex --project "$PWD"