CONTINUOUS · AGENTIC · STRUCTURAL

Hold the line.
Every commit.

Keelpin is the pin every line is belayed to. SAST, SCA, secrets, IaC, container, and pentest — fused into one canonical record per vulnerability and proven by an autonomous exploit before it reaches your inbox.

KP-001 / BELAYING PIN
HEMP CORDAGE Ø 0.625"
SHEAR LOAD
62,400 PSI
PIN HELD · 247 COMMITS
REV. 4.7.2
BUILD 3a8f1c
FORGED IN
AUSTIN, TX
SAST SCA · REACHABILITY WHITEBOX PENTESTER BLACKBOX PENTESTER BUSINESS LOGIC SECRETS SCANNING IAC SCANNING CONTAINER SCANNING CODE REMEDIATION CANONICAL FINDINGS JIRA SYNC SELF-HOSTED SAST SCA · REACHABILITY WHITEBOX PENTESTER BLACKBOX PENTESTER BUSINESS LOGIC SECRETS SCANNING IAC SCANNING CONTAINER SCANNING CODE REMEDIATION CANONICAL FINDINGS JIRA SYNC SELF-HOSTED
THE GAP

Your team ships code daily.
Your scanners disagree.

Most security platforms surface noise. Keelpin closes the gap between scattered scanner output and one canonical, exploit-validated truth. Every commit belayed to the same pin — not once a year.

01

Scanner sprawl

SAST says one thing. SCA says another. Your pentester said something else last March. Nobody trusts any of it.

02

The 364-day gap

Your team merges 200 PRs a week. Your annual pentest tested code that's already six months gone.

03

Unproven findings

Pattern-matchers flag the textbook patterns. Real exploits live in business logic, auth flows, and the seams between services.

THE PLATFORM

A fleet of pins.
One platform.

Continuous application security across every layer of your stack — from static analysis of your code to runtime pentesting of your apps. Each module is structurally part of the same pin.

PENTESTING

Hull

whitebox · code-aware

Agents read your source, model the architecture, and generate precise exploits validated against the live application.

Learn about Hull
PENTESTING

Tide

blackbox · runtime

Autonomous external pentesting against the running app. No code access. On-demand. Per repository. No subscription required.

Learn about Tide
CODE

Weld

agentic SAST

Code Property Graph plus LLM reasoning. Real vulnerabilities with full data-flow context — never regex matches, never blind hits.

Learn about Weld
LOGIC

Compass

business logic

Authorization bypass, IDOR, state-machine flaws, race conditions, and workflow abuse. The vulnerabilities pattern-matchers miss.

Learn about Compass
DEPENDENCIES

Cargo

SCA · reachability

Software composition analysis with reachability. Know which CVEs in your dependencies are actually reachable from attacker-controlled input.

Learn about Cargo
CREDENTIALS

Lockbox

secrets

Find leaked credentials, tokens, and API keys across code and commit history. Validated, deduplicated, prioritized by blast radius.

Learn about Lockbox
INFRA

Drydock

IaC scanning

Terraform, CloudFormation, Kubernetes manifests, and Helm charts — scanned for misconfigurations and policy violations before they sail.

Learn about Drydock
BUILD

Hold

container scanning

Scan container images for vulnerable packages, exposed secrets, and misconfigurations across every layer — before they ship to your registry.

Learn about Hold
THE PROCESS

From commit to verified weld.

Every Keelpin finding follows the same four-step process. Nothing is automated past the review gate. The pin holds because you decide it holds.

01

Scan

Push triggers SAST, SCA, secrets, IaC, and container scans across the changed surface. Source loads into ephemeral memory; nothing persists.

02

Prove

An agent generates an exploit and runs it against the live app. Confirmed exploits are filed as canonical findings with full reproduction.

03

Mend

You click a finding. An agent writes the patch and re-runs the original scanner. No patch is delivered unless the vulnerability is gone.

04

Hold

Patch lands as a clearly labeled bot PR in your normal workflow. You review. You merge. The pin holds. The finding closes.

REPORTING & ANALYTICS

One pin.
One source of truth.

Keelpin deduplicates SAST, SCA, secrets, IaC, container, and pentest results into a single canonical entry per vulnerability per repository — surfaced on a live dashboard and synced bidirectionally with Jira.

Canonical findings

Content-hash plus LLM semantic matching. One entry per vulnerability per repo, persistent across refactors and rebases.

Live dashboard

Risk, velocity, SLA, and MTTR trend charts. Drill into any repo, team, or severity. KPIs that map to how you actually ship.

Jira sync

One-click ticket creation. 15-minute status refresh. Hourly drift sweep on linked pairs. No manual reconciliation, ever.

~/keelpin/findings · juice-shop PIN HELD
ID FINDING SOURCE SEVERITY
KP-A38F JWT alg:none auth bypass HULL CRIT
KP-29B1 SQL injection on /rest/login WELD+HULL CRIT
KP-71C4 [email protected] RCE CARGO HIGH
KP-44E2 Hardcoded Alchemy API key LOCKBOX HIGH
KP-9A07 IDOR on /api/baskets/:id COMPASS HIGH
KP-1F58 Permissive CORS * WELD MED
KP-B321 Unpinned base image HOLD+DRYDOCK HELD
KEELPIN ENTERPRISE

Operates where
your hull lives.

Deploys entirely inside your AWS, GCP, or Azure account. Source, scan results, and AI inference stay inside your security perimeter. No managed control plane. No externally operated data plane.

Self-hosted

Run the entire platform inside your VPC. Fully air-gapped, no outbound calls required. Audited end-to-end on your infrastructure.

SSO & SCIM

SAML 2.0 or OIDC for sign-in. SCIM for automated user provisioning, deprovisioning, and group sync.

Deep integrations

GitHub, GitLab, Azure DevOps, Jira, Slack, plus Docker Hub, GHCR, Amazon ECR, and Google Artifact Registry.

DEPLOYMENT TOPOLOGY AIR-GAPPED
┌─ your-vpc.aws.amazon.com

├─ keelpin-control      ↪ kp-ctl.svc
├─ keelpin-scanner-pool  ↪ 8x kp-worker
├─ keelpin-pentest-pool  ↪ 4x kp-agent
├─ postgres + pgvector   ↪ findings store
├─ bedrock-runtime       ↪ your AWS account
└─ github-app            ↪ read-only by default

→ Outbound calls: 0
→ External egress: none
→ Source persistence: nil
TRUST & SECURITY

Your code stays
in your hold.

Keelpin is read-only by default. Source loads into ephemeral worker memory and is discarded when the scan completes. Only the canonical finding record persists.

Stateless processing

Source loads into ephemeral worker memory and is discarded when the scan completes. Nothing written to disk.

Never used for training

Zero-retention enforced upstream with every model vendor. No prompts, completions, or embeddings feed training pipelines.

Bring your own keys

Route inference through your own Anthropic, OpenAI, Bedrock, or self-hosted endpoint. Tokens never traverse Keelpin.

Read-only by default

Every scan runs on read-only scopes. The only feature that ever requests write access is Mend — opt-in per finding, never auto-applied.

Self-hosted available

Deploy the entire platform inside your VPC. Air-gapped, isolated, and audited end-to-end on your own infrastructure.

Only findings retained

After a scan, source is gone. Only the canonical finding record persists: rule, path, line, severity, status, and a redacted snippet.

COMPLIANCE-GRADE EVIDENCE

Pentest evidence accepted by every regime that requires it.

We don't report what might be vulnerable.
We hold the line on what is.

Schedule a structural review with the Keelpin team. We'll point Hull and Tide at a target you control, run a real exploit, and show you the canonical finding before the call ends.