Open source

ShakerScan Open Source

Run ShakerScan on your own infrastructure with the web UI, API, CLI, Scan, Hunt, findings, evidence, and the specialized security workflows: Continuous ASM, AI Gate and Model Intake (both preview), connected devices, and Fleet.

Only test systems you own or are explicitly authorized to assess. ShakerScan is a trusted-operator self-hosted product; keep the ordinary UI and API private.

release.facts
latest release
Version2.7.1
LicenseAGPL-3.0
Runs onLinux · macOS · Windows via WSL2
FleetOpt-in · Linux only
Web UIhttp://localhost:3000
APIhttp://localhost:8080
Sourcegithub.com/andriyze/shakerscan
What is included

The complete platform, not a community edition.

ShakerScan is an open-source security testing platform for web applications, APIs, AI systems, and network-connected devices. It runs locally in Docker and provides a web UI, REST API, CLI, persistent findings, evidence, repeatable Scan workflows, and AI-assisted Hunt workflows.

Scan

Deterministic web and API DAST: reviewed passive templates plus bounded XSS, SQL injection, and two-principal BOLA verification, with explicit active-testing authorization, budgets, findings, proof state, coverage, and retests.

Explore Scan

Hunt

Adaptive investigation planned by your Codex, Claude Code, or OpenCode session, with ShakerScan enforcing scope, capabilities, approvals, budgets, execution, and evidence.

Explore Hunt

Continuous ASM

Endpoint inventory, coverage gaps, bounded discovery and test waves, subdomain discovery, and certificate-transparency monitoring.

Explore Continuous ASM
Preview

AI Gate

Probe-pack testing for chat APIs, RAG endpoints, agent traces, MCP servers, and widgets, with transcripts, findings, and exports.

Explore AI Gate
Preview

Model Intake

Pre-deployment review of model artifacts: pinned provenance, dependencies, licenses, unsafe serialization, isolated runtime checks, and a signed, revocable admission.

Explore Model Intake

Connected devices

Inventory and posture for TVs, cameras, printers, routers, NAS systems, and appliances with coverage profiles, safety profiles, and a health circuit breaker.

Explore Connected devices

Fleet

Opt-in multi-node execution on Linux: one control plane places work on remote workers that connect over outbound-only HTTPS and return results centrally. A WireGuard transport is an operator preview.

Read the guide

Findings, evidence, and exports

One review surface across workflows: severity, proof state, source, lifecycle, retests, evidence export, HTTP transaction and HAR export, JSON and PDF reports.

Explore Findings, evidence, and exports
Install

One command. Then open localhost:3000.

The installer places ShakerScan under your home directory, creates the shakerscan command, starts the Docker stack, and binds the UI and API to local interfaces only. Clone the repository when you want to read or change the source. Docker and Docker Compose v2 are required; the first image pull can take several minutes.

start in three steps
Docker

Your first Scan

Install, open your agent, then paste the prompt.

  1. 1

    Install ShakerScan

    $ curl -fsSL https://install.shakerscan.com | sh
  2. 2

    Launch your coding agent

    $ shakerscan agent opencode

    also: claude · codex

    or skip the agent
    $ shakerscan start

    then open http://localhost:3000 for the dashboard

  3. 3

    Ask for a Scan, then a Hunt

    > Run a balanced passive Scan on https://app.example.test> Run a Hunt on this registered target.

Replace the example with a target you own or are authorized to test. Hunt needs the target registered first; Scan works without an AI agent.

Running ShakerScan on a remote VPS?

Tailscale

Use remote mode instead of the local start command. ShakerScan binds the UI and API to the VPS Tailscale address, saves the setting, and prints the private URLs to open from your laptop.

$ curl -fsSL https://install.shakerscan.com | SHAKERSCAN_REMOTE=1 sh
Never expose the scanner directly to the public internet.Remote/VPS guide

Remote VPS mode is not Fleet

Remote mode exposes one ShakerScan installation privately over Tailscale. Fleet connects multiple Linux worker machines to one control plane over outbound-only HTTPS. Never expose the ordinary UI or API directly to the public internet.

Client only: laptops, CI, and trusted LANs

The engine-less client needs Python 3.10+ and no Docker. It runs public posture checks, drives a remote engine, and serves MCP to your agent. LAN mode adds no authentication or encryption; use it on a trusted network only.

$ pipx install shakerscan# or: uv tool install shakerscan · brew install andriyze/shakerscan/shakerscan$ shakerscan check example.com# engine machine on a trusted LAN$ shakerscan start --lan# laptop$ shakerscan connect http://192.168.1.50:8080

In CI, submit with shakerscan scan --json and gate on GET /scans/{id}/deployment-decision; the scanner CLI can also write SARIF 2.1.

What’s new in 2.7.1

The latest release, from its notes.

ShakerScan 2.7.1, released October 6, 2026. Every release is built from a tagged commit, certified, and published with SBOMs and build attestations.

ImagesDocker Hub shakerscan/* · amd64 + arm64 · digests in RELEASES.md
SBOMsSPDX + CycloneDX · Sigstore-signed index · partial coverage
Buildsbuild attestations · Trivy high/critical gate · CodeQL
ClientPyPI trusted publishing
Full release notes for v2.7.1

Status: candidate prepared 2026-10-05. Publication and stable-channel promotion are pending.

This patch release fixes defects found after 2.7.0 shipped: in a post-release audit and while testing 2.7.0 behind an Enterprise gateway. It ships with client 0.7.5.

Proof and evidence

  • Authentication-bypass proof reads the whole response the transport accepted and returns no token signal when the JSON does not parse or part of the document lies beyond the bounded inspection walk. A rejection hidden past those limits can no longer become verified proof; the candidate stays an observation. Managed login is unchanged. (#327)
  • Purging a scan's HTTP archive covers the same child scans its export shows, clears their archive statistics and reports the owner ids it covered. Clearing a parallel or discovery run no longer reports 0 while leaving child traffic behind. (#327)
  • In-process scan capabilities are archived under their own capability name, and the archive separates external scanner traffic (relayed through an opaque pinned tunnel) from engine capabilities that recorded no calls. (#328)

Scans and findings

  • The scanner image bakes the nuclei template index, and workers rebuild it before taking jobs when it is missing. The first scan after a worker start no longer spends its passive-template budget indexing templates and finishing with no template findings. (#328)
  • A target with no published spec reports source_not_published instead of output_truncated. A scan that selects XSS or SQL injection but derives no candidates explains why, for example that JSON body APIs need state-changing HTTP permission. (#328)
  • Fix guidance is kept from the producing check and covers Git credential and configuration exposure, SSH password authentication and the connected-device checks. (#328)

Connected devices and Hunt

  • Connected-device scans and probes apply SHAKERSCAN_PRIVATE_NETWORK_TARGETS at submission and again in the device worker. The default still allows private networks; Lab environments keep their allowance. (#328)
  • A privileged Hunt start on a target whose scope is blocked reports the scope refusal and its reason first, instead of asking for an authorization receipt that could not be recorded. (#328)
  • Asset and Connected Devices views list confirmed open ports as discovered services; UDP probes with no response are shown separately as unconfirmed. (#328)

Install and upgrade

  • A root start after a non-root install moves the operator's files under results/ to the root API identity, so existing private archive payloads stay readable. Returning to a non-root start needs the manual chown described in docs/upgrade-and-rollback.md. (#327)

Delivery and validation

This note records the prepared scope, not a completed release certification. Exact-candidate certification, immutable version publication, public installer smoke and the stable-channel PR must succeed before the hosted installer advances.

Deployment model

A trusted-operator system on infrastructure you control.

Standalone ShakerScan runs on Linux, macOS, and Windows through WSL2. Project data, scan results, evidence, and artifacts are stored in your deployment, and there is no telemetry. It does pull images from Docker Hub, may fall back to public DNS-over-HTTPS for public names, and can call optional threat-intel services. It does not provide application login, users, roles, or tenant isolation; keep it behind a VPN, firewall, or authenticated reverse proxy. Optional Hunt and AI analysis use the agent or model provider you configure.

Operator UI / API / agents

One place to launch, schedule, review, and verify work.

Control plane

Scope, approvals, budgets, placement, findings, and evidence provenance.

Local workers

Default standalone execution.

Fleet workers (optional)

Outbound-only HTTPS, scoped credentials and job leases, no database credentials.

PostgreSQL · Redis queue · evidence · reports · artifacts, all in your deployment
Open source vs Enterprise

Open source is the platform. Enterprise adds organizational controls.

The self-hosted Enterprise edition is in public beta on top of the open-source release; managed hosting is in development. Nothing is sold as generally available yet; each row carries its current status.

CapabilityOpen sourceEnterprise
Complete platform: UI, API, CLI, Scan, Hunt, findings, evidenceIncludedIncluded (built on open source)
Continuous ASM, AI Gate (preview), Model Intake (preview), connected devices, FleetIncludedIncluded (built on open source)
Runs on infrastructure you controlIncludedSelf-hosted edition, public beta
Managed hosted environments—In development
OIDC single sign-on, roles, service tokens, audit—Public beta
Backup, restore, in-place upgrades, offline licence—Public beta
Support and pilot engagementsCommunity (GitHub issues)On request

Need SSO, organizational controls, or support?

Read the Enterprise page for current maturity labels, or talk to us about a pilot.

Explore Enterprise
Contribute and report

Issues, pull requests, and security reports are welcome.

Fork and clone the repository, create a feature branch, make and test the change, and open a pull request that describes the behavior and how you verified it. Report vulnerabilities through the security policy rather than a public issue.

Operator boundary

Keep the evidence where you can operate it.

ShakerScan does not claim multi-user authentication, tenant isolation, a hardened public SaaS boundary, guaranteed vulnerability discovery, or unrestricted autonomous exploitation. Review scope and authorization before every active test.