Install ShakerScan

See ShakerScan at work

A Scan and a Hunt,
in 50 seconds.

A real Scan. An agent-driven Hunt.
The evidence behind every decision.

Then see how it works

Recorded on our deliberately vulnerable lab. Real Scan and Hunt sessions, edited for time. In CI, GET /scans/{id}/deployment-decision returns the release-gate decision.

What the video shows

0:03 A new Scan of our lab target with a Thorough budget and active testing allowed. The real scan took seven minutes and is sped up.
0:15 The result: 22 findings confirmed with deterministic evidence and 9 still marked as needing verification, among them exposed phpinfo pages, private keys, .env files, backups and .git data.
0:21 The release gate blocks the deploy and lists the 22 blocking findings.
0:24 shakerscan agent opencode opens a coding agent, asked to hunt the same target for exposed secrets and verify what it finds. The agent proposes each step; ShakerScan scopes it, runs it and keeps the record.
The server's verifier re-checks each candidate: 17 findings are verified, and 3 stay unverified because the proof was inconclusive. Secret values are never shown to the agent.

Behind the evidence

How it works.

The big picture, then how a Hunt works.

01 / The big picture

Scan and Hunt
share one server.

Run a repeatable Scan or investigate with a Hunt. Both use the same server controls, registered capabilities, evidence, and target knowledge.

One execution foundation. Two ways to investigate.Open SVG

01

You set the boundaries.

Choose the target, objective, permissions, and limits.

02

ShakerScan runs the tests.

The server admits and meters each allowed action.

03

Evidence supports the result.

Deterministic proof controls verified findings.

02 / Agent-driven Hunt

Your agent plans.
The server executes.

Your terminal agent reads the evidence and chooses what to try next. ShakerScan checks the request, executes the allowed action, and returns the result.

An external terminal agent proposes a bounded action through MCP or CLI. The server checks target, policy, approval, and budget, executes it, and returns evidence for the next decision. An active planner drives the Hunt. — The agent plans here. ShakerScan executes there.Open SVG

Your agent

Observe. Question.
Choose the next action.

ShakerScan

Enforce. Execute.
Keep the evidence.

The proof boundary

AI proposes candidates.
Deterministic proof verifies.

An active external agent drives the investigation. Durable server state supports handoff and resume.

From the demo to the details

Try it on a target
you are authorized to test.

Install ShakerScan Connect your agent