.svg){kind=small}
Hey Siemba,
run a pentest.
Ask, and Siemba runs the test, scores the findings, and traces the attack chain, right inside your AI chat.
Siemba MCP is a Model Context Protocol server that connects Siemba's security platform to AI assistants. Ask a plain-English question in Claude, Cursor or any MCP client, and get finished security analysis back without opening a dashboard. No separate login and no context switch.
Requires a Siemba account with MFA.
Siemba MCP connects your assistant to Siemba. To security-test your own MCP servers, Talk to us →
Fix the auth check on /order/confirm and the whole chain collapses.
OAuth 2.1 + PKCE
Standard, proof-backed sign-in
MFA required
A code every session, no exceptions
Token rotation
Short-lived, refreshed automatically
Rate limited
Throttled so runs stay production-safe
Scoped to your account
Your roles, your assets, nothing else
.svg)
What can Siemba MCP do?
The MCP server gives your assistant the platform. Skills are the packaged workflows on top of it. Each one runs on AISO, Siemba's intelligence layer for test orchestration, insight-gathering and in-platform analysis. One command in plain English, one finished piece of security work. Siemba MCP ships with six skills at launch, and the set keeps growing.
Kill Chain
The Siemba MCP Kill Chain skill connects separate findings into the attack paths they really form, rates each by its weakest link, and names the single fix that collapses the most chains. One attack, not twelve findings.
You ask: "Are there any chained attack paths on acme.com?"
Remediation Plan
The Siemba MCP Remediation Plan skill groups every finding one change would fix, weighs the effort, then sequences by risk removed per unit of work. A finding list becomes a work list.
You ask: "What should we fix first?"
Critical Briefing
The Siemba MCP Critical Briefing skill puts every Critical and High finding across the estate in one read: what's most urgent, which assets carry it, and which fixes have quietly regressed.
You ask: "What are my most critical open findings right now?"
Attack Surface
Domains, certificates, TLS, registrars and geographies across everything exposed, plus the gaps the dashboards don't name: live but unmonitored, out of scope but still reachable, or a certificate about to expire.
You ask: "What do we have exposed to the internet?"
Test Analysis
The Siemba MCP Test Run & Analysis skill reads autonomous DAST run in whatever state it's in, down to the login screenshot when a run fails, since a bad credential or an unexpected MFA prompt is the usual culprit, and tells a broken scan apart from a target that blocked it, because the fixes differ.
You ask: "Why did last night's test fail?"
CISO Report
The Siemba MCP CISO Report skill produces a posture report that leads with direction of travel (improving, stable or worsening), then names the specific control each finding breaches across seven compliance frameworks instead of only counting them. Ready for the board, not the backlog.
You ask: "Generate a CISO report for the last quarter."
From a sentence to an answer, no console to open
Three moves cover most of it: ask for a test, read what came back, ask how it chains.
Say the target. Siemba takes it from there.
Name a target and Siemba maps it, picks the right test suite, and starts running, all automatically.
✓ test running · 12%
Severity, steps to reproduce, and the link back.
Findings come back scored, de-duplicated, with steps to reproduce and the compliance control each one violates.
One fix that breaks the whole path.
Ask which findings chain together and get the single fix that breaks the whole path.
Know what you own, and test behind the login
Pentesting is the headline move, but a good share of what's here is knowing what you own in the first place, and reaching what's behind a login.
Map your attack surface
Ask what's exposed and Siemba answers from the same live attack-surface data as the platform, down to individual asset dashboards and email security posture.
Test behind a login
Siemba validates saved credentials, including MFA-protected logins and VPN-gated environments, before a test starts, so an authenticated run doesn't fail halfway through.
Chat for speed. The platform for scale
This isn't a script wrapped around a scanner. Open-source pentest MCP servers hand your assistant one tool and a raw result. Siemba hands it a platform: authorization scoped to your account, findings de-duplicated and mapped to controls, signed audit-ready reports, and certified testers behind the same data. The chat is a faster way in, not a different, thinner product.
Fast, conversational, per asset
Best when you want an answer now: test this URL, show me what broke, tell me what to fix first, is this in scope.
- No context switch, no separate tool
- Uses your assistant's subscription, not ours
- Bring your own analysis on top of our findings
Portfolio, scheduling and evidence
Best when you're running a program: many assets, scheduled coverage, trend dashboards, and the signed report an auditor or customer will accept.
- CTEM dashboards and exposure trends
- Attestation letter and audit-ready reports
- Expert-led engagements and certified testers
How is Siemba MCP secured?
Handing an AI assistant the ability to run security tests only works if the platform holds the line, not the wording of a request. Here's what Siemba locks down by default.
You, not a service account
Every call runs as the user who signed in, with that account's roles and asset scope. No shared key, no over-provisioned token to leak.
Only your assets
Tests run against assets registered to your organization, under written authorization and agreed rules of engagement.
Production-safe by default
No denial of service, no stress testing, no brute forcing. Runs are throttled and adjustable mid-run.
Signed in, timed out, logged
OAuth 2.1 with PKCE and a required MFA code. Calls are rate limited and written to the platform activity log, so a test started from chat is as auditable as one from the UI.
How do I install Siemba MCP?
Setup takes minutes and works the same way across Claude, Cursor and other MCP clients.
Set up your Siemba account
You need an active Siemba subscription with MCP access and MFA enabled. See pricing →
Add Siemba as a connector
In Claude, an admin adds it once at the organization level and everyone signs in with their own account. Other clients take one config snippet, using the Siemba MCP server URL.
Start asking
Try list my recent autonomous DAST tests or show critical findings from the latest scan.
Don't take our word for it.
Ask the AI you already trust.
Fair enough, that's the whole point of the page. Copy the prompt, paste it into your assistant, let it read this page, and report back in its own words.
Questions we get asked
What is an MCP server?
Is this generally available or a preview?
What data leaves the Siemba platform when I use Siemba MCP?
Do I need a Siemba subscription?
Is Siemba MCP generally available?
Whose tokens does Siemba MCP consume?
How is a test started through Siemba MCP authorized?
Can it test anything I point it at?
Do I need to be a developer to set this up?
Does Siemba MCP cost extra?
Can Siemba MCP run in CI pipelines?
How is this different from open-source pentest MCP servers?
Is it safe to run against production?
Is a test started through Siemba MCP different from one started in the Siemba platform?
Does this replace the platform and the report?
My client says it could not attach to the server.
I signed in but the connection still failed.
How do I switch accounts, or fix an expired session?
What if your next pentest was just
a message away?
It already is. Connect your assistant, point it at one application, and watch a test run and a finding open in full. Thirty minutes, no slides.