DFTK — Digital Forensics Toolkit

DFTK is a Python toolkit for evidence-preserving digital-forensics operations. It provides structured results for files, archives, mobile artifacts, databases, captures, browser data, email, host artifacts, and timelines.

中文说明见 README.zh-CN.md.

Install

pip install dftk

Optional integrations:

pip install "dftk[email]"    # DKIM / SPF / DNS
pip install "dftk[ssh]"      # read-only SSH inventory
pip install "dftk[windows]"  # Registry / EVTX parsers
pip install "dftk[yara]"     # YARA rule scanning
pip install "dftk[mcp]"      # local MCP server
pip install "dftk[all]"      # all optional Python integrations

The core package has no mandatory third-party runtime dependencies. E01 filesystem traversal additionally requires pyewf / libewf bindings and pytsk3.

Start here

# Discover available capabilities
dftk list

# Build an Agent-ready intake manifest and next-step plan
dftk run evidence.intake --params '{"path":"/evidence/acquisition"}'

# Inspect one capability before running it
dftk describe artifact.inspect

# Analyze an artifact
dftk run artifact.inspect --params '{"path":"sample.apk"}'

# Save related observations in a case
dftk case --workspace /cases/intake new --name intake
dftk case --workspace /cases/intake run <case_id> artifact.inspect --params '{"path":"sample.apk"}'
dftk case --workspace /cases/intake export <case_id> --format md

Each run returns an Observation with a status, facts, evidence, warnings, and errors. unsupported, error, and blocked describe limitations or failures; they are not negative findings.

Agent and MCP use

For Agent use, the recommended entry point is this DFTK repository: give its URL to the Agent. It installs DFTK first, then runs a single bounded bootstrap that fetches the matching complete DFTK-skill bundle and emits a reviewable MCP configuration fragment:

dftk agent setup --root /evidence/acquisition --workspace /cases/intake --install-skill

See INSTALL_AGENT.md for the paste-ready instruction and AGENT_INTEGRATION.md for the complete operating loop.

DFTK includes a local stdio MCP server. Keep acquired evidence read-only and use a separate writable case workspace:

pip install "dftk[mcp]"
dftk mcp --root /evidence/acquisition --workspace /cases/intake --check
dftk mcp --root /evidence/acquisition --workspace /cases/intake

The server defaults to READ_ONLY with network access disabled. Its launch options define the evidence root, safety ceiling, network access, and timeout. See the MCP guide for configuration and policy details.

For an existing host configuration, install the matching Skill bundle directly:

dftk skill --install  # auto-detect the current Agent host; portable fallback: agents
# Inspect all supported target paths before a broad installation:
dftk skill --install --target all --dry-run

Documentation

• User guide — CLI, Python API, cases, observations, and audit logs.
• MCP guide — local server policy and host configuration.
• Capability map — domains and capability groups.
• Architecture — registry, evidence contract, and runtime boundaries.
• Safety policy — execution levels, network gates, and source-evidence protection.
• Toolchain deployment — external forensic binaries.
• Development guide — setup, tests, and contribution workflow.
• Agent reliability — bootstrap, coverage review, SQLite snapshot semantics, and executable validation.
• Documentation policy — ownership, generated data, and translation rules.

Sequence hunting: timeline.hunt_sequences correlates ordered events by shared entity fields, retaining source hashes, locators, and explicit coverage limits. See the sequence hunting guide for a runnable example.

Media evidence: consent-gated image/video previews, GPU-first OCR and derived evidence access.

Project facts

• One registry, 136 tools (131 READ_ONLY, 5 STATEFUL) — 121 primitives plus 15 recipes.

• Distribution: dftk; Python package: dftk; CLI: dftk.

• Python: CPython 3.10+.

• License: Apache-2.0.

• Maintainer: DyNooob · DigiForensics.

DFTK supports lawful, authorized examination of evidence. It is a technical toolkit, not legal advice.