Memex

Your Claude.ai chats and your Claude Code sessions live in two worlds that never talk. Memex makes them one memory you can search from both.

Local-first MCP server that indexes your Claude.ai chat history and your local Claude Code / terminal sessions, and exposes them to Claude Code (stdio) and Claude.ai (remote MCP connector).

Status: alpha, 0.3.1 on PyPI. Phases 0 to 7 closed, security-audited (plus four adversarial red-team rounds). Shipped: hybrid search, live capture, auto-summaries, chat ↔ repo association, the Claude.ai remote connector (GitHub OAuth), Claude Code / terminal ingestion with secret redaction, one-command setup, cross-platform autostart, and one-click claude.ai history backfill.

End-to-end demo: from Claude Code, you ask about something you discussed on claude.ai. Memex searches your chat history (search_chats) and Claude answers from the real conversation. No copy-paste, no manual context handoff.

Install

One command installs uv + Memex and runs memex setup (registers the MCP server, installs the always-on capture service, indexes your local sessions, prints the pairing token):

# macOS / Linux
curl -LsSf https://raw.githubusercontent.com/dioniipereyraa/memex/main/scripts/install-pypi.sh | sh

# Windows (PowerShell)
powershell -ExecutionPolicy ByPass -c "irm https://raw.githubusercontent.com/dioniipereyraa/memex/main/scripts/install-pypi.ps1 | iex"

Then the one browser step a terminal cannot do: install the Chrome extension, paste the token it printed, and click "Backfill claude.ai history". Other paths (pipx, uv tool install, from source) are under Installation.

Related MCP server: conversation-history-mcp

The problem

Brainstorming and planning happen in Claude.ai. Execution happens in Claude Code. The two worlds do not talk to each other: Claude Code cannot read a chat of yours from Claude.ai, not even the one that originated the task it is currently working on. The memory Anthropic shipped on Claude.ai (March 2026) is curated, not full history, and lives isolated inside Claude.ai.

Memex fills that gap: runs locally, indexes the entire corpus of your chats, and exposes them as MCP tools so Claude can search and pull past context whenever it needs to.

How it works

[Claude.ai]
    ↓  (official JSON export / Chrome ext)
[Ingestor]  →  [SQLite + sqlite-vec]  →  [local embeddings (fastembed / Ollama)]
                                    ↓
                          [core: storage + retrieval]
                                    ↓
                  [MCP stdio]  ───→  Claude Code, Claude Desktop
                  [MCP Streamable HTTP] ─→ Claude.ai connector

Design: pure core (storage, ingest, embeddings, retrieval) decoupled from transport. The same engine serves both stdio and remote MCP without a rewrite.

Installation

Quick install (one command)

Installs uv if needed (it brings the right Python), installs Memex from PyPI, and runs memex setup to wire it into Claude Code (MCP server + always-on capture service + local session indexing + the pairing token). No git clone required.

# macOS / Linux
curl -LsSf https://raw.githubusercontent.com/dioniipereyraa/memex/main/scripts/install-pypi.sh | sh

# Windows (PowerShell)
powershell -ExecutionPolicy ByPass -c "irm https://raw.githubusercontent.com/dioniipereyraa/memex/main/scripts/install-pypi.ps1 | iex"

The only step a terminal cannot do is the browser one: install the Chrome extension, paste the token the installer prints, and click "Backfill claude.ai history" to import your history. Embeddings work out of the box (a quantized 130 MB fastembed model downloads on first ingest; no Ollama, no API key).

A plain pipx install memex-chats (or uv tool install memex-chats) followed by memex setup does the same thing in two steps. The autostart service is generated for the wheel install too (launchd / systemd / a logon Scheduled Task), so no clone is needed for it.

From source (for development)

Clone the repo to get an editable install plus the extension source and service templates. You need git; the script installs uv and the pinned Python for you.

macOS / Linux

git clone https://github.com/dioniipereyraa/memex
cd memex
./scripts/install.sh

install.sh installs uv if you do not have it, runs uv sync, and verifies the install with memex doctor. That is the whole setup.

Windows (PowerShell)

git clone https://github.com/dioniipereyraa/memex
cd memex
powershell -ExecutionPolicy Bypass -File .\scripts\install.ps1

Same three steps as the macOS script (install uv, sync, verify). If git is not installed, get it from git-scm.com first.

Manual (any OS, if you prefer explicit steps)

# 1. Install uv (skip if you already have it):
#    macOS/Linux:  curl -LsSf https://astral.sh/uv/install.sh | sh
#    Windows:      powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
git clone https://github.com/dioniipereyraa/memex
cd memex
uv sync               # installs Python 3.13 + dependencies into .venv
uv run memex doctor   # verify

Upgrading (already have memex)

Upgrade with the same tool that installed it, or the new version will not land on your PATH. A previous install owns the memex executable, so installing again with a different manager fails with "Executables already exist" and memex keeps running the old version. (This is the usual reason memex sync seems to "not exist" after an upgrade: an old pipx-owned memex shadowed a newer uv tool install.)

# 1. Find which tool owns it:
which memex            # macOS / Linux
Get-Command memex      # Windows (PowerShell)

# 2. Upgrade with that tool:
pipx upgrade memex-chats        # if pipx owns it
uv tool upgrade memex-chats     # if uv owns it

If you want to switch managers, uninstall the old one first (pipx uninstall memex-chats or uv tool uninstall memex-chats), then reinstall. After upgrading, memex --version shows the new version and memex doctor should pass.

Multi-device (optional)

To make Claude one memory across several machines, install Tailscale first (so each device has a stable private address), then opt in once per device:

memex setup --sync     # opt-in, persisted: the always-on service is sync-reachable

It prints the single memex sync connect ... line to run on your other devices. The choice is saved, so the service stays sync-reachable across reboots without re-running anything. Full flow, security model, and the manual one-shot alternative are in Syncing across your devices.

For a dual-boot machine (Linux and Windows that are never on at the same time), use file-based sync instead: memex setup --sync-dir <shared-folder> on each OS, pointed at the same folder. No Tailscale, no two devices online at once. See Dual-boot or never-online-together.

First run

The fast path is one command:

uv run memex setup

memex setup wires everything end to end and is safe to re-run: it registers the MCP server with Claude Code (claude mcp add), installs the always-on live-capture service (launchd on macOS, systemd on Linux, a Scheduled Task on Windows), indexes your local Claude Code sessions, and prints the access token to paste into the Chrome extension. Each step degrades to a warning instead of aborting the rest. Flags: --no-mcp, --no-autostart, --no-ingest, --remote (also install the claude.ai connector agent), -y (no prompt).

After that, the only manual step left is installing the Chrome extension and pasting the token, so new chats are captured as you go.

Manual steps (if you prefer to run them yourself)

# 1. Request your official Claude.ai export (Settings → Privacy → Export data),
#    then ingest it (first run downloads the embedding model, ~30s):
uv run memex ingest /path/to/your-export.zip

# 2. Optionally index your local Claude Code / terminal sessions:
uv run memex ingest-claude-code

# 3. Search, or wire it into Claude Code (see "Wiring it into Claude Code"):
uv run memex search "your query" -n 5
uv run memex stats

(If you installed with a script, the commands above work as written. If you installed the published PyPI package instead, drop the uv run prefix.)

Where your data lives. From a cloned repo, the database and exports stay in <repo>/data. From a uvx/pip install, they go to your OS data directory (~/Library/Application Support/memex on macOS, %LOCALAPPDATA%\memex on Windows, ~/.local/share/memex elsewhere). Override with MEMEX_DB_PATH / MEMEX_EXPORTS_DIR. memex doctor prints the resolved path.

Using Memex (after setup)

Once memex setup finishes, nothing else needs launching: the capture server and the indexers run in the background. You use Memex three ways.

1. From Claude Code (the main way). Setup registers the MCP server, so Claude Code can search your whole history and keep it fresh through these tools:

Just ask Claude Code naturally ("what did we decide about X in my claude.ai chats?", "index this session", "sync this to my other devices") and it calls these. index_terminal_sessions and sync_now are local-only (never exposed to the remote claude.ai connector). Note sync_now sends what is already indexed, so to include the session you are in, ask Claude to index_terminal_sessions first. To have context injected automatically at the start of a session, add the optional SessionStart hook.

2. From the terminal. memex search "...", memex stats, memex doctor, memex token (re-print the extension pairing token), memex --version.

3. Capture (automatic, you run nothing). The Chrome extension indexes a claude.ai chat as you open or reload it (and its "Backfill" button imports your old history once); your Claude Code sessions are indexed when a session ends, plus a 15-minute backstop.

How fresh is the data (real-time)?

Same-machine is essentially real-time. Cross-device is not a live relay (it is peer-to-peer with no cloud): paired devices auto-reconcile every 15 minutes while both are online, and you can force it immediately with memex sync now. A claude.ai chat is captured by the extension on open/reload, so to sync the latest turns of an in-progress chat, reload it (so the extension re-captures it), then run memex sync now. The terminal cannot read claude.ai directly, so the extension is always what captures; sync only propagates what is already indexed locally.

Embeddings backend

Default is fastembed (zero-config). To route through a local Ollama instead (e.g. you already run it for other models):

export MEMEX_EMBED_BACKEND=ollama
ollama pull nomic-embed-text          # install Ollama first: https://ollama.com/download

On Windows, install Ollama from ollama.com/download before setting MEMEX_EMBED_BACKEND=ollama. This is optional; fastembed needs none of it.

Diagnostics

Run memex doctor any time something is not working. It checks Python version, database, embedder, live-capture server, summarizer config, registered repos, and indexed corpus. Reports OK / WARN / FAIL per check.

Troubleshooting

Start with memex doctor (above): most of the issues below also surface there.

memex: command not found right after installing. The memex executable lives in the tool's bin directory, which may not be on your PATH in the current shell yet. Open a new terminal, or add it now:

• macOS / Linux: export PATH="$HOME/.local/bin:$PATH"

• Windows: re-open the terminal so the installer's PATH change takes effect (the bin dir is %USERPROFILE%\.local\bin for uv tool, or the pipx Scripts dir).

Then confirm the right binary: which memex (macOS/Linux) or where memex (Windows). It should point at the uv-tool / pipx install, not an old clone.

The extension shows the server as not responding, or ingest returns 401.

• Is the server up? curl -s http://127.0.0.1:5777/health should return JSON. If it does not, start it (memex serve) or check the autostart service below.

• 401 means the pairing token is missing or stale: run memex token, copy the value, paste it into the extension popup, and click Save token.

The autostart service did not come up. Check its status, then read its log:

• Linux: systemctl --user status memex-serve; log at ~/.local/share/memex/serve.log. Failed to connect to bus means there is no user systemd manager for this session (common over plain SSH): run loginctl enable-linger "$USER", make sure XDG_RUNTIME_DIR=/run/user/$(id -u) is set, then re-run memex install-service.

• macOS: launchctl list | grep memex; log in the data dir (~/Library/Application Support/memex/com.memex.serve.log on a PyPI install, <repo>/data/serve.log from source).

• Windows: Task Scheduler, or schtasks /Query /TN MemexServe /V; log at %LOCALAPPDATA%\memex\serve.log. If an older install left a task with the same name, it can silently mask the new one: delete it (schtasks /Delete /TN MemexServe /F) and re-run memex install-service.

Autostart does not survive logout or reboot (Linux). A systemd user unit is torn down when your session ends unless lingering is enabled: loginctl enable-linger "$USER".

The one-command installer fails with a shell error. The installer is POSIX sh. If you are on a very old or unusual shell and the pipe fails, fetch and run it with bash instead: curl -LsSf <url>/install-pypi.sh | bash.

Windows: uv tool install fails with "Acceso denegado (os error 5)" copying memex-mcp.exe. Two causes, usually together:

• Claude Code / Claude Desktop is running and its MCP child process holds the running .exe. Close Claude Code and Claude Desktop, then retry.

• Windows Defender blocks writing a brand-new .exe even when nothing holds it. The tell: the destination file does not exist yet (Test-Path "$env:USERPROFILE\.local\bin\memex-mcp.exe" is False) but the copy still fails. Add a Defender exclusion for %USERPROFILE%\.local\bin and %APPDATA%\uv.

Then re-run with uv tool install --force memex-chats (or uv tool install --reinstall --refresh <spec> when updating from a branch).

Windows: a paired device times out reaching your sync port (5777). In sync-reachable mode the always-on server binds 0.0.0.0:5777, but Windows Firewall silently drops inbound connections (the peer just times out, never "connection refused") until a rule allows the port. memex setup --sync adds it for you (idempotent, best-effort); if setup was not elevated it prints the one-time admin command to run yourself: New-NetFirewallRule -DisplayName "Memex sync 5777" -Direction Inbound -Protocol TCP -LocalPort 5777 -Action Allow.

Sync times out even though tailscale ping works. If tailscale ping <peer> succeeds but regular ping/TCP (and memex sync) time out in BOTH directions, the problem is a local TUN/routing conflict on one of the machines, not memex and not the firewall rule: another VPN-style adapter (Hamachi, another VPN, a virtual switch) is swallowing the OS traffic that should ride the Tailscale interface. Disable or remove the other adapter and restart the Tailscale service. (Seen live: a Hamachi adapter installed for game hosting broke a Windows box's Tailscale traffic while tailscale ping kept working.)

The first backfill (or first ingest) is slow. The first embed downloads a ~130 MB quantized model; that is one-time, later runs are fast. The backfill is incremental and safe to re-run, so a slow or interrupted first pass simply resumes where it left off.

Ingest uses too much memory. Peak RAM scales with the embed batch size. Lower it: set MEMEX_EMBED_BATCH_SIZE=1 (about 0.67 GB peak) in your environment or .env.

Where is my data? memex doctor prints the DB path. By default it is ~/.local/share/memex (Linux), ~/Library/Application Support/memex (macOS), or %LOCALAPPDATA%\memex (Windows) for a PyPI install, or <repo>/data from a clone. Override with MEMEX_DB_PATH / MEMEX_EXPORTS_DIR.

MCP server tools (v1)

• search_chats(query, limit=5, source?, mode="hybrid", repo?) searches the corpus. Modes: hybrid (default, combines vector search + FTS5 BM25 via Reciprocal Rank Fusion), semantic (vectors only), lexical (FTS5 only, ideal for proper nouns or exact terms). source filters by origin (conversations, design_chat, memory). repo boosts results associated to a registered repo (see "Repo associations" below). Deduplicated per conversation.

• get_chat(uuid, messages_limit=10, messages_offset=0) fetches a conversation with its messages, paginated. raw_content is omitted; each message is truncated to 1500 chars to stay inside the client's token budget (worst-case response ~17k chars). Long chats are paginated with messages_offset; max messages_limit is 100.

• list_recent_chats(limit=10, source?) lists the latest chats ordered by last update.

• find_related(context, limit=5, repo?) takes free-form text (a paragraph, a file contents, the current discussion) and returns chats that are semantically related. Pure vector search, no FTS, capped at 4000 input chars. Useful when you want "more like this" without typing a keyword query.

The stdio server (memex-mcp, used by Claude Code) also exposes two write/maintenance tools, so you can ask Claude to keep your memory fresh instead of dropping to a terminal. They are local-only: never registered on the remote claude.ai connector (they mutate the store and reach the network).

• index_terminal_sessions() runs the incremental ingest-claude-code scan (indexes new/changed local Claude Code sessions, including the current one up to what is flushed to its transcript). Shares the single-flight ingest lock; returns counts.

• sync_now() runs an immediate two-way reconcile with every paired device and, if a shared folder is configured, a file sync. Requires sync enabled. It propagates what is ALREADY indexed, so to include the current session call index_terminal_sessions first.

Search is also reachable from the CLI with memex search "query" --mode {hybrid|semantic|lexical}. For databases created before the hybrid FTS5 work, run memex reindex-fts once to populate the lexical index.

Wiring it into Claude Code

Once your local database is populated (memex ingest), start the MCP server with uv run memex-mcp. For Claude Code to discover it automatically, add a .mcp.json file at the root of your project (or a user-level server in ~/.claude.json):

{
  "mcpServers": {
    "memex": {
      "command": "uv",
      "args": ["run", "memex-mcp"],
      "cwd": "/absolute/path/to/the/memex/repo"
    }
  }
}

Set cwd to the absolute path where you cloned Memex (where pyproject.toml lives). Restart Claude Code and the tools search_chats, get_chat, list_recent_chats will show up in the session.

The same searches are also available from the CLI via uv run memex search "..." if you prefer them outside Claude Code.

Connecting from claude.ai (remote MCP, Phase 4)

memex serve-remote exposes the same 4 tools as a custom connector for claude.ai. One connector works across claude.ai web, Claude Desktop, and the mobile apps: the connection is brokered by your Claude account and always originates from Anthropic's cloud, never from your device. That has two consequences:

1. The server must be reachable on a public HTTPS URL (a hostname with a public IPv4 A record; localhost and private IPs are rejected by design). The intended setup is a tunnel that terminates TLS and proxies to Memex on loopback.

2. Auth is either none or full OAuth 2.0 (claude.ai has no field to paste a token). Memex uses OAuth backed by a GitHub OAuth App, restricted to an allow-list of GitHub usernames: anyone else who completes the OAuth dance still gets a 401 on every request. Your machine must be on (and the tunnel up) for the connector to respond.

1. Publish a URL with Tailscale Funnel

Install Tailscale, log in, then:

tailscale funnel --bg 8377

This prints your public URL, e.g. https://my-mac.my-tailnet.ts.net, TLS included. Any other tunnel works the same way (ngrok, Cloudflare Tunnel); Memex only needs the resulting https:// URL.

2. Create a GitHub OAuth App

At github.com/settings/developers > OAuth Apps > New OAuth App:

• Homepage URL: your funnel URL.

• Authorization callback URL: <funnel URL>/auth/callback.

Copy the Client ID and generate a Client Secret.

3. Configure and run

In .env (see .env.example):

MEMEX_REMOTE_BASE_URL=https://my-mac.my-tailnet.ts.net
MEMEX_GITHUB_CLIENT_ID=Iv1.xxxxxxxxxxxx
MEMEX_GITHUB_CLIENT_SECRET=xxxxxxxxxxxxxxxxxxxx
MEMEX_REMOTE_ALLOWED_GITHUB_LOGINS=your-github-username

uv run memex serve-remote   # listens on 127.0.0.1:8377, endpoint at /mcp

The server refuses to start with incomplete config (no insecure defaults: an empty allow-list would expose your whole chat history to any GitHub account).

4. Add the connector in claude.ai

Settings > Connectors > Add custom connector, with URL https://my-mac.my-tailnet.ts.net/mcp. claude.ai registers itself via dynamic client registration, sends you through the GitHub authorization, and the tools appear in your chats. OAuth state is persisted encrypted on disk, so restarting serve-remote does not break the connection.

Security model in short: loopback bind + tunnel, OAuth proxy over GitHub with a username allow-list enforced on every request (revoking the app on GitHub locks it out immediately), TrustedHostMiddleware pinned to the public hostname, and the same indirect-prompt-injection envelope on tool results as the stdio transport.

Indexing Claude Code / terminal sessions (Phase 6)

Memex also indexes your local Claude Code and terminal sessions, so a single search covers both halves of your history: what you discussed on claude.ai and what you built with Claude Code. Claude Code records each session as a JSONL file under ~/.claude/projects/; Memex reads them directly (no extension, no network). This covers Claude Code wherever it runs, the terminal CLI and the VS Code / JetBrains extensions alike: they all write the same transcripts.

uv run memex ingest-claude-code        # scans ~/.claude/projects/**/*.jsonl
uv run memex ingest-claude-code --path /custom/path   # alternate root

These sessions are stored under the claude_code source, searchable like any other chat (and filterable with source="claude_code"). Details:

• Incremental. Unchanged sessions are skipped (by content hash), so re-running after more work is cheap. Sessions grow append-only, so a re-scan picks up new turns and new sessions. Re-run it whenever you want to refresh, or wire it to a schedule.

• Repo-aware for free. Every session carries its working directory, so each is auto-associated with the registered repo of that cwd (run memex repos add <path> first). search_chats(repo=...) then boosts the sessions where you worked on that project.

• What is indexed: your prompts, the assistant replies, and tool calls as [tool_use: ...] / [result] markers. Excluded: assistant internal reasoning (thinking), parallel sub-agent side threads, and CLI plumbing (slash-command echoes, bash wrappers). Everything stays in the local SQLite DB.

• Secret redaction. Session logs capture real terminal output and file contents, which often contain credentials. Before storing, Memex masks common secret shapes (API keys, bearer/JWT tokens, PEM private keys, KEY=/SECRET= assignments, URLs with embedded passwords) as [REDACTED:...], and does not persist the raw block content for this source. Best-effort, not a guarantee: it catches well-known formats so third-party credentials that were never really part of a conversation stay out of the index (this matters because the remote connector can surface indexed text to claude.ai).

Getting a Claude Code session indexed

A session becomes searchable once its transcript is scanned. Day to day there are three ways to trigger that yourself:

• Ask Claude. Say "index this session" and Claude calls the index_terminal_sessions MCP tool (the current conversation is captured up to what Claude Code has flushed to its transcript at that point). Follow with "sync my devices" (sync_now) if you want it on your other machines right away.

• Close the conversation. With the SessionEnd hook below installed, ending the Claude Code chat (just the conversation, not the whole terminal) ingests that session the moment it closes.

• Run the command. memex ingest-claude-code scans everything new right now.

Even if you do none of these, the periodic backstop below picks up new and grown sessions within about 15 minutes of their last write. The two automatic mechanisms complement each other (both run in the background at low priority, and the embedding model only loads when there is something new, so they barely cost anything):

1. SessionEnd hook — ingests a session the moment it closes (no delay, nothing lost). Add to ~/.claude/settings.json:

   {
     "hooks": {
       "SessionEnd": [
         { "matcher": "*", "hooks": [
           { "type": "command", "command": "/absolute/path/to/memex/scripts/session-end-hook.sh" }
         ] }
       ]
     }
   }

   The hook reads the closed session's transcript_path and ingests just that one session, detached, returning immediately (SessionEnd hooks cannot block Claude Code).

2. Periodic backstop — a scheduled scan that catches anything the hook missed (e.g. a crash). On macOS, with launchd:

   sed "s|__REPO__|$(pwd)|g" scripts/com.memex.ingest-claude-code.plist.template \
     > ~/Library/LaunchAgents/com.memex.ingest-claude-code.plist
   launchctl load ~/Library/LaunchAgents/com.memex.ingest-claude-code.plist

   It runs scripts/scheduled-ingest.sh every 15 minutes as a low-priority background job. A PyPI install does not need these manual steps on macOS or Windows: memex setup (or memex install-service) already schedules the backstop there (the com.memex.ingest-claude-code launchd agent / the MemexIngest Scheduled Task). On Linux the installer sets up only the serve, so wire the backstop yourself with a systemd user timer or cron running memex ingest-claude-code.

   Per-OS state of the two mechanisms: the SessionEnd hook script is bash, so it works on macOS and Linux; on Windows it needs a PowerShell equivalent of session-end-hook.sh, not yet included, so there the ways in are the backstop task, asking Claude, or the manual command.

Syncing across your devices

If you run memex on more than one machine (say a laptop and a desktop), memex sync makes Claude one memory across them: index a chat on one device and search it from the other. Each device keeps its own local store and they sync peer-to-peer, with no central server and nothing leaving your hardware. The sync needs both devices powered on and reachable at the same time (it is overlap sync, not a cloud relay). It ships in memex 0.4.0+; on an older install, upgrade first (uv tool upgrade memex-chats).

The whole feature is off by default and exposes nothing until you turn it on. While it is off, the sync endpoints return 404 and the sync commands refuse, so a normal single-device install has no extra surface.

Set it up once (recommended)

Install Tailscale on each device first, then opt in once:

memex setup --sync

This is the normal setup plus one thing: it turns sync on and makes your always-on capture service sync-reachable on this device's Tailscale address (it binds 0.0.0.0 so one server still answers the local extension, with the Host allow-list pinned to loopback + your Tailscale IP, resolved at startup, and the per-install token gating access). The choice is persisted: set it once and the service comes up sync-reachable on every reboot, so you never hand-run a serve command. It prints the single line to run on each other device:

memex sync connect --url http://100.x.y.z:5777 --token <TOKEN> --name my-mac

connect turns sync on there, pairs, and runs a two-way reconcile, so that device is set up and caught up in one step.

Set-and-forget: memex setup --sync also enables auto-sync, so paired devices reconcile by themselves every 15 minutes while both are online (no env var, no cron). It is overlap sync, not a cloud relay: if the other device is off, that tick is skipped and catches up next time both are on. On top of the timer, a capture also triggers a sync a few seconds after it lands: when the Chrome extension captures a chat and Memex indexes it, that new conversation is pushed to your other devices right away (no need to run anything). A burst of captures (a backfill, several open tabs) is coalesced into one sync once it settles, and it only fires while both devices are on (for a dual-boot the other OS just picks up the change on its next boot). Turn just this off with MEMEX_SYNC_ON_CAPTURE=false; tune the coalescing window with MEMEX_SYNC_ON_CAPTURE_DEBOUNCE_SECONDS (default 10). You can still sync immediately by hand between ticks (e.g. to push the chat you are having right now), with:

memex sync now          # immediate two-way reconcile with every paired device

memex sync now propagates what is already indexed locally; a claude.ai chat is indexed when the extension captures it (on open/reload), so reload an in-progress chat first if you want its latest turns included. memex sync status shows what is paired and the last sync; memex sync reconcile is the same as now. Turn the whole thing back off (loopback-only, sync + auto-sync off) with memex setup --no-sync. Tune the interval with MEMEX_SYNC_INTERVAL_SECONDS (default 900).

One-shot (without changing your setup)

If you would rather not make the service permanently reachable, start a temporary endpoint on the device that has the conversations:

# SOURCE device:
memex sync serve
# -> Memex sync serve reachable at http://100.x.y.z:5777
#    On the other device, run this one command:
#      memex sync connect --url http://100.x.y.z:5777 --token <TOKEN> --name my-mac

memex sync serve works the same on macOS, Linux, and Windows (it uses your Tailscale address, so there is no per-OS network setup and no --host/allow-list juggling). If you are not on Tailscale, pass --host <address the other device can reach>. It binds to that address only and runs alongside your loopback capture server without a port clash; keep it up while the other device connects, then stop it with Ctrl+C. Afterwards, memex sync reconcile re-syncs an already-paired device.

reconcile is the usual command: it syncs both ways and converges the two devices, keeping the newer copy of anything that exists on both (last writer wins by update time). If the same conversation was edited on both devices with the exact same timestamp, that is a conflict: reconcile leaves both copies untouched and reports it, and you pick a winner with a one-directional memex sync pull --peer mac (take the peer's) or push --peer mac (send yours).

Every sync only transfers conversations that are new or changed, brings their embeddings along so your machine never re-embeds, and is idempotent (re-running moves nothing new). A large history transfers in size-bounded batches automatically, so it works no matter how many conversations you have (tune the target with MEMEX_SYNC_MAX_BATCH_BYTES, default 8 MB; a single conversation larger than the peer's MEMEX_INGEST_MAX_BODY_BYTES is skipped with a clear message). Both devices must use the same embedding model (the sync refuses on a mismatch). The peer's token is stored user-only (0600) next to the DB. Pairing is a trust decision (it shares an access token), so only pair devices you control. Never sync the SQLite file itself with a folder-sync tool; always use memex sync, which is consistent at the conversation level.

Auto-sync (the every-15-minutes reconcile) is turned on for you by memex setup --sync; you do not need an env var. Under the hood it reconciles with each paired peer on startup and every MEMEX_SYNC_INTERVAL_SECONDS (default 900), skipping a peer that is offline and backing off while an ingest is running. If you did not use setup --sync (e.g. a manual/dev setup) you can still enable it with MEMEX_SYNC_AUTO=true before memex serve; either the persisted flag or the env var starts the loop. It only runs while sync is enabled; turn the whole feature back off with memex sync disable (or memex setup --no-sync).

Dual-boot or never-online-together (file-based sync)

Network sync needs both devices online at the same time. That never happens on a dual-boot machine (Linux and Windows on the same disk, only one running at a time), and is awkward for any pair of devices that are rarely on together. For that, memex has a second sync mode that uses a shared folder instead of a live connection: each OS exports a snapshot of its store to the folder and imports the others'. No server, no port, no token, no third device. It ships in memex 0.4.4+.

Point every OS at the same shared folder (one command per OS):

# On Linux (the Windows partition mounted at, e.g., /mnt/windows):
memex setup --sync-dir /mnt/windows/memex-sync --device-name linux

# On Windows (the SAME physical folder, its path there):
memex setup --sync-dir C:\memex-sync --device-name windows

That is it. From then on each OS, when it boots, auto-syncs through the folder (every 15 minutes while running), and memex sync now syncs immediately. The folder is created for you if it does not exist, and the choice is persisted, so you set it up once.

Where to put the folder on a dual-boot: it must be readable from both OSes, so put it on the Windows / NTFS partition. Linux reads and writes NTFS out of the box; Windows cannot read Linux's ext4. (For two separate machines that are rarely on together, any shared location works: a USB drive, or a cloud-synced folder like Dropbox.)

Dual-boot gotcha: Windows Fast Startup (and hibernation) leaves the NTFS partition dirty, so Linux mounts it read-only, and file sync needs to write its snapshot there (the initial export fails loudly if it cannot). If the folder mounts read-only on Linux, disable Fast Startup on Windows (or powercfg /h off) and shut Windows down fully once.

How it converges: a device only ever writes its own <device>.memexsync.gz and reads the others', so they never clash. Sync is last-writer-wins by update time, exactly like the network mode, and a same-timestamp conflict (a fork) is left untouched and reported. The other OS's newer conversations land on your next boot (it left its snapshot on disk), and yours reach it on its next boot. Give the two OSes distinct --device-name values if their hostnames are the same, or they would overwrite each other's snapshot. The snapshot carries your conversation text and vectors, so keep the shared folder somewhere only you can read (the same "only share what you control" rule as pairing a peer). The two modes are independent: you can use file sync, network sync, or both.

Rotating a token

Each device has one access token (the ingest_token file next to its database; memex doctor shows the folder). Rotate it whenever it was pasted somewhere it should not live: a chat, a screenshot, a ticket. There is no rotate command yet, and the running server reads the file only once, so the order matters:

1. Delete the token file (rm <data-dir>/ingest_token; PowerShell: Remove-Item -Force $env:LOCALAPPDATA\memex\ingest_token).

2. Restart the always-on serve (macOS: launchctl unload then load on com.memex.serve; Linux: systemctl --user restart memex-serve; Windows: schtasks /End /TN MemexServe then schtasks /Run /TN MemexServe). Skipping this leaves the OLD token accepted (it is cached in memory) and the new file ignored.

3. Run memex token to see the new value. Re-paste it into that machine's Chrome extension popup, and re-pair every device that dials this one: memex sync connect --url http://<this-device>:5777 --name <name> --token "..." (passing --token avoids the hidden prompt, which is easy to mistype).

Running always-on

memex serve (claude.ai capture) is a long-lived server, and the Claude Code ingest backstop runs on a schedule. To have them start on login and stay up (restarting if they crash), let memex setup install the service, or call it directly on any OS:

memex install-service            # install (launchd / systemd / Scheduled Task)
memex install-service status     # show what is registered
memex install-service uninstall  # remove it

This is cross-platform: launchd agents on macOS, a systemd user unit on Linux, a Scheduled Task on Windows. By default it manages the live-capture server plus the 15-minute Claude Code ingest backstop. Idle cost is low: neither loads the embedding model until there is real work. It works the same from a PyPI / pipx / uv tool install (a self-contained launchd / systemd / Scheduled Task that runs the installed memex serve directly) and from a cloned repo. Use a persistent install (pipx or uv tool), not transient uvx, so the service has a stable interpreter to launch at boot.

To also run the claude.ai remote connector as a service, pass --remote (macOS): memex install-service --remote. It only makes sense once the MEMEX_REMOTE_* config is in .env and the Tailscale Funnel is up on the same port (tailscale funnel --bg 8377), otherwise the agent crash-loops; that is why it is opt-in.

cd /path/to/memex
for svc in serve ingest-claude-code; do
  sed "s|__REPO__|$(pwd)|g" "scripts/com.memex.$svc.plist.template" \
    > ~/Library/LaunchAgents/com.memex.$svc.plist
  launchctl load ~/Library/LaunchAgents/com.memex.$svc.plist
done
launchctl list | grep memex

To stop a service: launchctl unload ~/Library/LaunchAgents/com.memex.<name>.plist.

Auto-summaries (Phase 3, optional)

When search_chats returns a chat that does not have a summary yet, Memex can generate one on-the-fly using Claude Haiku. The summary is persisted, so the next search of the same chat hits cache and does not pay the API again. This way you only pay for chats you actually look at, not for the whole corpus.

Opt-in. Off by default. Cap: at most 3 summaries generated per search_chats call (parallel), to bound latency and per-query cost.

1. Install the extra (adds the anthropic SDK):

   uv sync --extra summaries

2. In your .env (or as env vars), set:

   ANTHROPIC_API_KEY=sk-ant-...
   MEMEX_SUMMARY_ENABLED=true

3. Use search_chats from Claude Code as usual. The first time a chat appears in results without a cached summary, Memex generates one. Subsequent searches return the cached summary instantly.

Configurable via env: MEMEX_SUMMARY_MODEL (default claude-haiku-4-5-20251001), MEMEX_SUMMARY_MAX_TOKENS (default 200). If the API fails for a particular chat (no key, rate limit, network), that one result comes back without a summary, the search itself never aborts, and the warning is logged.

Cost model: bulk ingest of an export with 74 chats costs $0 (no summaries are generated at ingest time). Each unique chat you actually open in a search costs roughly $0.01 (Haiku, ~5-10k input + ~200 output tokens). $5 of API credits comfortably covers months of use.

Repo associations (Phase 3)

Memex can associate each chat with the local code repos it touches, and boost those chats when search_chats is invoked from inside a repo. So when you ask Claude Code something like "remember the auth refactor we discussed?", chats that touched the current repo rank higher than unrelated chats with the same keywords.

How it works:

1. Register a repo:

   uv run memex repos add /path/to/your/repo

   Memex reads .git/config (origin URL), and pyproject.toml / package.json / Cargo.toml (package name). The canonical key prefers the git remote (stable across clones) over the path.

2. Run a one-time scan over chats you already ingested:

   uv run memex repos scan

   Each chat gets matched against every registered repo. The matcher uses 4 signals (highest wins): remote URL literal in chat text (confidence 1.0), absolute path literal (0.9), manifest name word-bounded (0.8), display name word-bounded (0.5). Anything below 0.5 is dropped. New chats ingested after registering the repo are auto-scanned at ingest time.

3. From Claude Code, search with the repo argument:

   search_chats(query="auth refactor", repo="d:/dionisio/memex")

   Or pass the git remote URL, or the canonical key from memex repos list. Chats associated to the repo get their distance reduced by 0.3 * confidence. Chats outside the repo still appear lower down (it is a boost, not a filter).

CLI helpers:

uv run memex repos list           # show registered repos
uv run memex repos remove <key>   # unregister + cascade-remove associations
uv run memex tag <chat-uuid> <repo-key>    # manual override (sticky vs auto-scan)
uv run memex untag <chat-uuid> <repo-key>  # remove an association

A manual tag survives subsequent auto-scans: once you tag a chat by hand, the matcher will not overwrite it.

Proactive context injection (SessionStart hook)

Claude Code supports hooks that run shell commands at session boundaries. Memex ships memex session-context, designed to be wired into the SessionStart hook so every new Claude Code session in a registered repo starts with a Markdown blob listing the most relevant past chats. No query needed.

To enable it, add to your .claude/settings.json (project-local) or ~/.claude/settings.json (user-global):

{
  "hooks": {
    "SessionStart": [{
      "command": "uv run memex session-context"
    }]
  }
}

The command auto-detects the active repo by walking up from cwd until it finds .git. If the repo is registered in Memex and has associated chats, it prints a short Markdown blob with the top N (default 5) chats, ordered by manual first, then auto by confidence. If anything fails (no .git, repo not registered, no associations), it prints nothing on stdout (only diagnostics on stderr), so the hook is a silent no-op in that case.

You can also run the command by hand to debug: uv run memex session-context [--repo <path>] [--limit N].

Live capture (Phase 2)

So that new Claude.ai chats land in Memex without asking for a manual export:

1. Start the local HTTP server in a terminal:

   uv run memex serve

   Listens on 127.0.0.1:5777 by default. Keep it running while you browse claude.ai. On start it prints an access token; the extension must send it (the loopback Origin check alone does not authenticate other local processes). Reprint it any time with uv run memex token.

2. Load the Chrome extension:

   • Soon: install from the Chrome Web Store (in review for the alpha).

   • For now (unpacked load):

     • Open chrome://extensions/

     • Enable Developer mode

     • Load unpacked → pick chrome-extension/ from the repo

     • Click the Memex icon and confirm the "Server" chip says responding (green).

3. Pair the extension with the access token. Run uv run memex token, copy the value, paste it into the extension popup's token field, and click Save token. This is a one-time step per machine (the token lives user-only next to your DB). Without it, ingest requests are rejected with 401.

4. Import your full history (one click). With a claude.ai tab open, click Backfill claude.ai history in the popup. It pages through your entire chat list and pulls every conversation into Memex, skipping the ones already indexed and unchanged (so re-running is cheap and a closed tab just resumes on the next click). Progress shows in the popup. This is what makes a fresh install searchable without a manual export.

5. Then just use claude.ai. Every new chat you open or create is ingested automatically. Verify with memex stats or by calling search_chats from Claude Code.

Details in chrome-extension/README.md.

Autostart (macOS + Windows + Linux)

So you do not have to run memex serve by hand every time you log in (this is also what memex setup does for you):

memex install-service          # default action: install
memex install-service status   # check current state
memex install-service uninstall

The CLI dispatches to the right installer for your OS:

• macOS: writes launchd agents for serve + the 15-minute Claude Code ingest backstop, and launchctl loads them. Logs in the data dir (<repo>/data/serve.log from a clone, ~/Library/Application Support/memex/com.memex.serve.log from a PyPI install). Add --remote to also run the claude.ai connector (needs the MEMEX_REMOTE_* config). See Running always-on.

• Windows: registers a Scheduled Task (MemexServe) that runs memex serve at logon. No admin required, no console window, survives VS Code close. Logs at %LOCALAPPDATA%\memex\serve.log. Auto-restarts up to 3 times if it dies.

• Linux: writes a systemd user unit at ~/.config/systemd/user/memex-serve.service, enables it, and starts it now. Logs at ~/.local/share/memex/serve.log. To keep it running across logout: loginctl enable-linger $USER. Status: systemctl --user status memex-serve.

From a pip/pipx install (no cloned repo), memex install-service works on macOS, Linux, and Windows alike: it generates a self-contained agent (launchd / systemd / a logon Scheduled Task) that runs the installed memex serve directly, with the database in your per-user data directory. Use pipx (not transient uvx) so the service has a stable interpreter to launch at boot.

Making Claude use Memex proactively

By default, LLMs are conservative with tools: they prefer to ask before invoking anything. If you say "remember we talked about X?", Claude tends to answer "I don't recall" instead of searching.

The docstrings of the 4 tools already include "USE PROACTIVELY" instructions, but you can reinforce it by adding this snippet to your CLAUDE.md (global at ~/.claude/CLAUDE.md for every session, or local at <project>/CLAUDE.md for a specific one):

## Memex: persistent memory of Claude.ai chats

There is an MCP server `memex` with 4 tools: `search_chats`, `get_chat`, `list_recent_chats`, `find_related`.
They index ALL of the user's Claude.ai history, reachable via hybrid search
(semantic + lexical FTS5).

**Operational rule:** before answering "I have no record", "I don't remember", "this is
the first time I hear about this", or anything equivalent, call `mcp__memex__search_chats`
with the relevant query. Claude Code's native memory starts clean every session; Memex
is the only path into the user's real history.

Typical triggers: "remember when...", "did I tell you about...", "we already talked about...",
"the other day we discussed...", or any reference to a project / person / decision that might
live in history.

Roadmap

See ROADMAP.md for phases, close criteria, and current status (in Spanish, it is the internal journal).

Devlog

See DEVLOG.md for the log of decisions, blockers, and progress (in Spanish).

Inspiration and references

• Official feature request: anthropics/claude-code#12858

• Claude Historian, claude-conversation-extractor: MCP servers for Claude Code / Desktop history. Reference for tool structure.

• claude-conversation-export: Claude.ai exporter using the same capture strategy. Useful as backfill.

• Spin-off of the SyncChat project.

License

MIT.