Skip to main content
Glama
keyuraghao

edb-explorer

by keyuraghao

Why

Evidence lives in databases: the Active Directory store, SRUM's per-app network usage, a phone's messages and call log, browser history, macOS's knowledgeC, a web app's LevelDB cache, an old Access application, a mysqldump from a compromised server. Each has its own file format, its own timestamp encoding and its own "where is the interesting table" knowledge. Most tools handle one of them, on one OS.

EDB Explorer opens all of them on Linux, Windows and macOS with pure-Python parsers (no esent.dll, no sqlite3 locks on the evidence, no native LevelDB), decodes timestamps per application, and exposes the same read-only engine three ways:

Interface

Command

For

Desktop GUI

EDB-Explorer / edb-explorer

analysts browsing many databases side by side

CLI

edb-explorer info / tables / dump / sql / views / timeline / export / report …

scripting and quick triage

MCP server

edb-explorer mcp

AI agents (Claude Desktop, Claude Code, Cursor, any MCP client)

Everything is strictly read-only - the tool never writes to an evidence file.

Related MCP server: Access and SQLite MCP Server

Features

Any database, one UI

  • ESE / JET Blue, SQLite (WAL sidecars merged, SQLCipher-aware parser), LevelDB directories, Access .mdb/.accdb, dBase .dbf (incl. soft-deleted records), Berkeley DB, mysqldump / pg_dump / sqlite .dump text, mongodump .bson, Windows Event Logs .evtx (Security / System / Application / Sysmon, one table per channel).

  • Formats are detected by signature, not extension; scan a folder or a mounted image and everything supported is listed.

  • Open many databases at once (Ctrl+O), drag-and-drop files or folders, welcome screen with Open / Recent / Scan buttons, a Tasks panel showing a progress bar for every file being opened or table being loaded while you keep working, collapse / expand all databases with one click.

  • No table is too big: every row of every table is loaded, whatever the file size. The first rows of a table (250,000 by default) stay in memory; a bigger table continues into a temporary SQLite cache on disk and the grid, sorting, filtering, the inspector and the extract dialog all work on the full row set from there, so memory stays flat. Settings ▸ Preferences… (Ctrl+,) explains the trade-off and sets the budget with a slider / exact field (presets, live RAM estimate, Restore default) and the cache directory (free space shown; EDB_EXPLORER_CACHE_DIR works too).

  • Light and dark themes (View ▸ Theme: Light / Dark / Follow system, or the sun / moon button at the right of the toolbar, Ctrl+Shift+D); every toolbar action has its own icon.

Application knowledge (see docs/formats.md)

  • 57 profiles: NTDS, SRUM, Exchange, WebCache, Windows Search, UAL, Windows Timeline, Notifications, Windows Security / System / Application / Sysmon event logs, Chrome/Edge history, cookies, logins, autofill, Firefox, Safari, iOS Messages, AddressBook, CallHistory, knowledgeC, Photos, Notes, Calendar, backup Manifest, WhatsApp (iOS + Android), Android contacts / calls / SMS / downloads / media, Telegram, Skype, Signal Desktop, Thunderbird, Zeitgeist, macOS Quarantine / TCC / Notification Center, Chromium Local Storage / IndexedDB, WordPress dumps, RPM databases …

  • Friendly table names, per-column timestamp decoding (WebKit, Cocoa s/ns, Unix s/ms/µs, FILETIME, OLE, PRTime) and artifact views - ready-made SQL such as Browsing history, Messages with handles, Network usage by application, Quarantined downloads - one double-click away.

Exchange mailbox viewer (Ctrl+M, or double-click Mailboxes under an Exchange database)

  • Mail-client layout: mailboxes → folders → message list → preview (rendered HTML / plain text / internet headers / recipients / attachments / every MAPI property). Works on Exchange 2013+ databases by decoding the store's property blobs directly - no Exchange server, no PST.

  • Export selected messages, a folder or a whole mailbox as EML (attachments embedded), HTML, TXT or JSON; save attachments; export the message list to xlsx/csv/pdf.

(No screenshot of the mailbox viewer: the only databases it opens are real mailboxes, which are case data.)

AI agents inside the app (Ctrl+Shift+A)

  • Embedded terminal running Claude Code, OpenAI Codex CLI, Gemini CLI, GitHub Copilot CLI, Aider, a shell or any command. Configure MCP registers this tool's MCP server with the agent (scoped to the folders of the open databases), Login / Start do the rest - the agent can then call run_sql, timeline, exchange_messages, generate_report… on the evidence you have open.

Analysis

  • SQL console over any format: tables are materialised into SQLite with decoded values; every open database is attached as a schema so you can join a phone's messages against a laptop's browser history.

  • Timeline: every timestamp column of every table (detected by profile hints or value heuristics) becomes an event; filter by date range, double-click to jump to the record, extract to xlsx/csv/pdf.

  • Column statistics: nulls, distinct values, min/max/mean, top values, and the detected timestamp encoding + date range per column.

  • Record inspector with hex dump and every interpretation of a value (UTF-16, SID, GUID, 9 timestamp readings).

  • Find in databases (Ctrl+Shift+F): substring or regex across every column of every table of every open database; double-click a hit to jump to the record.

Extract (Ctrl+E, right-click → Extract selected rows…)

  • Selected rows, displayed rows, query results, whole tables or whole databases → CSV, XLSX, JSON, JSON Lines, TXT, PDF (a database → one workbook with a sheet per table).

Reports (Ctrl+R) - file metadata, SHA-256, format state, table inventory with counts, schema and sample rows, case ID / analyst / notes → HTML, PDF, DOCX, Markdown, XLSX, TXT, JSON.

Project files (File ▸ Export project… / Import project…, Ctrl+Shift+P / Ctrl+Shift+I, or drop a .edbproj on the window) - share exactly what you have open with a colleague on another computer:

  • The project records every open database (path, size, SHA-256), the open tabs with their filters and hidden columns, the SQL console text, the dock layout and your notes; optionally it embeds the evidence files so it is self-contained (files of any size stream in 4 MiB chunks).

  • Integrity: every member of the file is listed in a SHA-256 hash map that is verified before anything is used; on import each evidence file is checked against its recorded hash, whether it was embedded or found on the other machine (original path → same relative path under a folder you point at → same name + hash).

  • Signing: an Ed25519 key pair is created for you on first use; the hash map is signed, and the importer shows the signer and their key fingerprint, keeps a trust store of fingerprints you have accepted, and flags anything altered after signing.

  • Confidentiality: a password encrypts the manifest and the embedded files with AES-256-GCM (scrypt-derived key); the signature stays verifiable without the password. Format details in docs/projects.md.

Supported databases

Format

Examples

Notes

Microsoft ESE

ntds.dit, SRUDB.dat, Exchange .edb, WebCacheV01.dat, Windows.edb, UAL .mdb, DataStore.edb

schema, indexes, long values, multi-values, template tables; Exchange mailboxes decoded into a mail-client view

SQLite 3

iOS sms.db, AddressBook.sqlitedb, CallHistory.storedata, knowledgeC.db, Photos.sqlite, Android contacts2.db / mmssms.db / calllog.db, WhatsApp, Chrome History / Cookies / Login Data, Firefox places.sqlite, Safari History.db, ActivitiesCache.db, wpndatabase.db, TCC.db

WAL merged, no locks on evidence

LevelDB

Chrome/Edge Local Storage, Session Storage, IndexedDB, Discord, Teams, Slack, VS Code

live and all_records (incl. deleted / superseded)

Access

.mdb (Jet 3/4), .accdb (ACE)

dBase / FoxPro

.dbf (+ .dbt/.fpt memos)

soft-deleted records exposed

Berkeley DB

RPM Packages, cert8.db / key3.db, sendmail maps

btree / hash / recno

SQL dump

mysqldump, pg_dump (incl. COPY), sqlite .dump

loaded into SQLite, then queryable

BSON dump

mongodump collections

nested documents flattened to JSON

Windows Event Log

Security.evtx, System.evtx, Application.evtx, Microsoft-Windows-Sysmon%4Operational.evtx

one table per channel; System fields + flattened EventData/UserData columns; logon / privilege / process views

The full list of profiles, their signature tables and views is in docs/formats.md.

Install

Binaries (no Python needed)

Grab the build for your OS from the releases page:

OS

Full install

Portable

Windows

EDB-Explorer-<ver>-setup.exe - Start Menu / desktop shortcuts, optional .edb association and edb-explorer on PATH. Uninstall from Apps & features (or the Start Menu entry); the uninstaller offers to remove your settings, signing key and caches too.

EDB-Explorer-<ver>-windows-x64-portable.zip - unzip anywhere, run EDB-Explorer.exe. Settings, keys and caches stay in its data folder (nothing touches the registry or your profile). Uninstall with Uninstall.cmd or by deleting the folder.

Linux

edb-explorer_<ver>_amd64.deb - sudo apt install ./edb-explorer_<ver>_amd64.deb, remove with sudo apt remove edb-explorer. Or edb-explorer-<ver>-linux-x86_64.tar.gz - extract, ./install.sh adds a menu entry and ~/.local/bin/edb-explorer; ./uninstall.sh removes it (optionally with your data and the folder).

EDB-Explorer-<ver>-linux-x86_64-portable.tar.gz - extract, run ./EDB-Explorer; everything lives in data. Uninstall = delete the folder.

macOS (Apple Silicon)

EDB-Explorer-<ver>-macos-arm64.dmg - drag to Applications; unsigned, so right-click ▸ Open on first launch (or xattr -dr com.apple.quarantine). The disk image also carries Uninstall EDB Explorer.command.

EDB-Explorer-<ver>-macos-arm64-portable.zip - unzip, run the app from the folder; state stays in data.

A portable.txt file next to the executable is what switches the app into portable mode; the installed variants never write there. Every checksum is published as a .sha256 file next to the download.

Each bundle also contains the edb-explorer console program for the CLI and MCP server. Running it without arguments opens the GUI too.

From PyPI / source

pip install "edb-explorer[all]"          # GUI + MCP
pip install "edb-explorer[mcp]"          # headless: CLI + MCP server only (no Qt)

# or from a checkout
git clone https://github.com/keyuraghao/EDB_FILE_EXPLORER && cd EDB_FILE_EXPLORER
uv sync --all-extras && uv run edb-explorer gui

Requires Python 3.10+. On minimal Linux servers the GUI needs the usual Qt runtime libraries (libegl1 libgl1 libxkbcommon0 libdbus-1-3 libfontconfig1).

Usage

GUI

edb-explorer                                       # welcome screen (Open files / Open recent / Scan folder)
edb-explorer gui SRUDB.dat sms.db "Local Storage/leveldb" places.sqlite

Shortcut

Action

Ctrl+O / Ctrl+Shift+O

open files / scan a folder

Ctrl+Q / Ctrl+L / Ctrl+I

SQL console / timeline / column statistics

Ctrl+F / Ctrl+Shift+F

filter rows in the current table / search every table

Ctrl+E / Ctrl+Shift+E

extract table or view / extract selected rows

Ctrl+R

generate report

Ctrl+T

timestamp decoder

Ctrl+Shift+D

toggle light / dark theme

Ctrl+W / Ctrl+Shift+W

close tab / close database

Ctrl+, / Ctrl+Shift+K

preferences / keyboard shortcuts

These are the defaults. Settings ▸ Keyboard shortcuts… (Ctrl+Shift+K, also under Help) lists every menu action grouped by menu with its current and default key; select one, press the new combination and Assign (or Enter). Conflicts are flagged and can take the key over from the other action; Clear, Reset to default and Reset all are one click, Copy as text gives a cheat-sheet, and the bindings persist across sessions (shortcuts/<action id> in the settings). Panel toggles and theme choices can be given shortcuts too.

CLI

edb-explorer info SRUDB.dat --hash                     # metadata, profile, SHA-256
edb-explorer tables ntds.dit --count                   # table inventory with record counts
edb-explorer schema SRUDB.dat "Network Data Usage"     # columns and indexes (friendly names work)
edb-explorer dump SRUDB.dat SruDbIdMapTable -n 20 -g svchost -f jsonl
edb-explorer search WebCacheV01.dat "login.microsoftonline.com" --regex
edb-explorer export ntds.dit -o out/ -f xlsx           # whole database -> out/ntds.xlsx
edb-explorer export SRUDB.dat -t "Network Data Usage" -o net.pdf -f pdf
edb-explorer report SRUDB.dat ntds.dit -o case.docx --case CASE-2026-001 --analyst "J. Doe"
edb-explorer scan /mnt/evidence                        # every supported database, by signature
edb-explorer sql sms.db "SELECT COUNT(*) FROM message WHERE is_from_me = 1"
edb-explorer views History                             # list artifact views for this database type
edb-explorer views History history -o history.xlsx     # run one and extract it
edb-explorer stats places.sqlite moz_places            # column statistics + timestamp encodings
edb-explorer timeline sms.db History SRUDB.dat -o timeline.xlsx --start 2021-01-01
edb-explorer summary knowledgeC.db                     # analysis overview (JSON)
edb-explorer formats                                   # what can be opened
edb-explorer timestamp 132565120200137766              # FILETIME? OLE? Unix? Cocoa? all readings
edb-explorer mailboxes "Mailbox Database.edb"          # Exchange: mailboxes
edb-explorer mail "Mailbox Database.edb" -m 129 --folder Inbox --export out/ -f eml
edb-explorer agents --configure claude --allow /cases/001   # register the MCP server with Claude Code
edb-explorer project export case.edbproj ntds.dit Security.evtx --embed --ask-password   # signed, encrypted, self-contained
edb-explorer project info case.edbproj                 # verify hash map + signature, list contents
edb-explorer project import case.edbproj --dest ~/cases --ask-password --gui   # unpack, verify, open
edb-explorer project trust                             # your signing fingerprint + trusted signers

Every command has --json output where it makes sense and --help.

MCP server (AI agents)

edb-explorer mcp                                  # stdio (default) - what Claude Desktop / Claude Code / Cursor use
edb-explorer mcp --allow /cases/001/evidence      # restrict what the agent may open (recommended)
edb-explorer mcp --transport streamable-http --port 8765
edb-explorer mcp SRUDB.dat ntds.dit               # pre-open databases at startup

Claude Code:

claude mcp add edb-explorer -- edb-explorer mcp --allow /cases/001/evidence

Claude Desktop / Cursor (claude_desktop_config.json, .cursor/mcp.json):

{
  "mcpServers": {
    "edb-explorer": {
      "command": "edb-explorer",
      "args": ["mcp", "--allow", "/cases/001/evidence"]
    }
  }
}

Docker (headless, Streamable HTTP on :8765, evidence mounted read-only):

docker run --rm -p 8765:8765 -v /cases/001/evidence:/evidence:ro ghcr.io/keyuraghao/EDB_FILE_EXPLORER

Tools exposed (31) - open_database, scan_directory, list_databases, close_database, get_database_info, database_summary, list_tables, describe_table, count_records, query_table (paging + filters), get_record, get_record_raw (hex + every decoding), search, run_sql (any format, cross-database), list_views / run_view, column_statistics, detect_timestamps, timeline, export_table_to_file, export_database_to_directory, generate_report, interpret_timestamp, list_known_profiles, list_formats, exchange_mailboxes / exchange_folders / exchange_messages / exchange_message / exchange_export / exchange_save_attachment; resources edb://databases, edb://{db}/tables, edb://{db}/{table}/schema; prompt triage_database. See docs/mcp.md.

An agent conversation typically looks like:

"Open the phone's sms.db and the laptop's Chrome History, tell me what happened on 2021-01-30 and write a report."open_database ×2 → database_summaryrun_view("sms", "messages")timeline(start="2021-01-30", end="2021-01-31")run_sql(...)generate_report(...)

Architecture

src/edb_explorer/
├── core/                 UI-agnostic engine (no Qt)
│   ├── backends/         one module per format: ese, sqlite, leveldb, access, dbf, bsddb, sqldump, bsondump, evtx
│   │   └── __init__.py   signature detection + registry
│   ├── database.py       Database: thread-safe read-only wrapper over a backend
│   ├── session.py        many open databases, stable ids, path allow-list, directory scan
│   ├── profiles.py       ESE profiles + Profile/ArtifactView model;  profiles_apps.py: phones, browsers, macOS, Windows, servers
│   ├── values.py         timestamp / SID / GUID / UTF-16 decoding, encoding guesser, hexdump
│   ├── sqlworkspace.py   SQL over any backend (materialised into SQLite, one schema per database)
│   ├── analysis.py       timestamp detection, column statistics, timeline, summary
│   ├── exchange/         Exchange store: ProP property blobs, RTF, mailboxes/folders/messages, EML/HTML export
│   ├── rowstore.py       disk-backed row store so grids can hold tables of any size (SQLite cache, sort/filter views)
│   ├── project.py        .edbproj project files: SHA-256 hash map, Ed25519 signing, AES-256-GCM encryption
│   ├── agents.py         AI agent CLI detection and MCP registration (Claude Code, Codex, Gemini, Copilot)
│   ├── search.py, formats.py, export.py, report.py, updates.py
├── gui/                  PySide6 application (welcome page, task panel, mailbox viewer, embedded terminal, dialogs,
│                         rebindable shortcuts, preferences)
├── mcp/                  MCP server built on the core
├── portable.py           portable mode (portable.txt beside the executable -> state in ./data)
└── cli.py                Typer CLI

Adding a format = one backend module implementing tables() / iter_rows() / info() plus a signature in backends/__init__.py; adding an application = one Profile in profiles_apps.py.

from edb_explorer.core import Session
from edb_explorer.core.sqlworkspace import SqlWorkspace
from edb_explorer.core.analysis import build_timeline

s = Session()
phone = s.open("sms.db")             # SQLite, profile "iOS / macOS Messages"
laptop = s.open("History")           # SQLite, profile "Chromium History"
srum = s.open("SRUDB.dat")           # ESE, profile "SRUM"

ws = SqlWorkspace()
print(ws.run_view(phone, "messages", limit=5).rows)
print(ws.query_database(srum, "SELECT COUNT(*) FROM SruDbIdMapTable").rows)
events, _ = build_timeline([phone, laptop, srum], limit=10_000)

Development

uv sync --all-extras
uv run pytest                                     # 100+ unit tests on fake / synthetic fixtures, no evidence needed
EDB_EXPLORER_TEST_DB=/path/to/any.edb uv run pytest -m integration
uv run ruff check src tests && uv run ruff format src tests && uv run mypy src/edb_explorer/core
./packaging/build.sh                              # Linux/macOS bundles, .deb, portable   (Windows: packaging\build.ps1)
python scripts/make_demo_data.py                  # synthetic evidence folder (History, sms.db, Security.evtx)
python scripts/make_screenshots.py                # regenerate docs/screenshots from that synthetic data

Screenshots and demo data never come from real evidence: scripts/make_demo_data.py generates deterministic fake databases and scripts/make_screenshots.py drives the real GUI offscreen with a neutral user and host name.

See CONTRIBUTING.md for the layout, how to add a database profile and the release process, and SECURITY.md for MCP hardening notes.

Acknowledgements

License

MIT - see LICENSE.

Related MCP Connectors

Related MCP Servers

  • A
    license
    Not graded
    quality
    D
    maintenance
    Enables AI to interact with Microsoft Access (.mdb, .accdb) and SQLite 3 databases, including read-only support for Access 97 databases. Supports executing SQL queries, importing/export data from CSV and Excel files, and storing notes about database files.
    MIT
  • A
    license
    Not graded
    quality
    D
    maintenance
    Enables AI interaction with Microsoft Access and SQLite databases to perform SQL queries, updates, and data management. It supports importing and exporting data via CSV and Excel files and includes tools for managing human-readable notes about database files.
    7
    MIT
  • A
    license
    A
    quality
    B
    maintenance
    Lets AI agents query local SQLite database files read-only using Node's built-in sqlite module, providing tools for listing tables, describing schemas, and running SQL queries.
    3
    5 npm
    MIT
  • A
    license
    Not graded
    quality
    C
    maintenance
    Enables exploring and querying SQLite databases through natural language, with tools to list tables, describe table structures, and run SELECT queries.
    MIT