About OmniSentient

A security platform that
earns its trust in code.

Built because we got tired of manual dependency upgrades after every CVE. OmniSentient collapses vulnerability detection, AI root-cause isolation, and PR remediation into a single non-disruptive, human-approved control plane.

View Architecture Specs → Read Trust Center
Architecture Proof Verified System Spec
Ed25519 Platform Key
Asymmetric provenance signatures on all forensic exports.
SHA-256 Hash Chain
PostgreSQL DB-trigger enforced append-only audit ledger.
Zero-Write Direct Commit
All fixes arrive strictly as PRs requiring human signoff.
PostgreSQL RLS FORCE
Strict tenant isolation on all forensic event tables.
Fail-Closed RBAC
Viewer role mutations blocked at API layer with 403.
Our Mission

Trust as a structural property — not a marketing claim.

In early 2025, every engineering team we knew was handling Dependabot alerts the exact same way: someone would notice an alert, manually open a PR, wait for CI build feedback, and merge. Sometimes that process was delayed for weeks; sometimes it was skipped entirely.

Most security tools ask for broad write access to your entire codebase, then ship policy documents promising they will behave responsibly. OmniSentient takes the opposite approach: read what's necessary, write nothing without explicit human approval, and record every action before it executes.

Every platform guarantee is observable in code review, in runtime behavior, and in the audit trail itself. Engineering teams can deploy our remediation pipeline without negotiating a new security review every time an update ships — because the structural boundaries protecting their code on day one remain enforced indefinitely.

Product Evolution

How the platform has evolved.

From a single GitHub webhook handler to a full forensic control plane. Per-release technical breakdowns are documented in our changelog.

Origin

Initial Webhook Handler

First automated feedback loop — capturing failing CI runs and posting automated diagnostic comments in pilot repositories.

Inspect Implementation Proof →
Original webhook receiver deployed via Flask route at /webhook with GitHub App HMAC-SHA256 signature verification.
Remediation

Dependabot & Remediation PR Pipeline

Automated Dependabot alert ingestion connected directly to human-approved Pull Request remediation workflows.

Inspect Implementation Proof →
Interactive review queue at /dashboard/queue with human approval/discard buttons and zero auto-merge.
Multi-Tenant

Supabase Multi-Tenant Control Plane

Strict Row-Level Security (RLS) policies and multi-tenant GitHub App installation tracking for enterprise pilots.

Inspect Implementation Proof →
PostgreSQL FORCE ROW LEVEL SECURITY on forensic tables to prevent cross-tenant data leakage.
Governance

Quota Management & Billing Controls

Stripe integration, organizational quota enforcement, and initial role-based access control (RBAC) tiers.

Inspect Implementation Proof →
Tiered quota management wired to /dashboard/quota and billing webhook handlers.
Phase 1 Ship

Institutional Frontend & Telemetry HUD

Full design token system rollout, sovereign sidebar navigation, real-time telemetry HUD, SBOM generation, and supply chain controls.

Inspect Implementation Proof →
System HUD and SBOM pipeline live at /dashboard/dependencies and /dashboard/sentinel.
Institutional v1.1

Incident Command & Ed25519 Hash Chaining

Incident Command System, Ed25519 cryptographic signatures on forensic exports, PostgreSQL DB-level forward hash chaining, and Sentinel hotspot ranking.

Inspect Implementation Proof →
Audit ledger verification at /security-audit and Ed25519 key signing specs in SECURITY_POLICY.md.
Hardening

Server-Authoritative RBAC & AuditService

Forensic query layer (`AuditService`), server-enforced 403 API protection for viewer roles, canonical cross-surface deep linking, and live DB persistence.

Inspect Implementation Proof →
auth_guard.py middleware enforcing HTTP 403 forbidden responses on state-mutating viewer calls.
Operating Principles

Four non-negotiable ideas we refuse to compromise on.

Enforced at the infrastructure layer, validated by automated tests, and observable in the audit ledger.

PRINCIPLE 1

Zero-Write

OmniSentient never commits code directly to main branches. All remediation arrives as a Pull Request and requires explicit human approval before merging.

Inspect Approval Queue →
PRINCIPLE 2

Manifests by Default

Left alone, the pipeline reads dependency manifests and advisory metadata. Source files are read when you start a scan or index a repository, and written only to open a fix pull request — which is what the App’s contents: write permission is for.

View System Specs →
PRINCIPLE 3

Fail-Closed Authorization

Any authorization anomaly defaults to denial. Database timeouts, malformed JWTs, or missing claims resolve to HTTP 401/403—never to privilege escalation.

Read Security Policy →
PRINCIPLE 4

Non-Repudiable Audit

Every state transition logs UTC timestamp, actor identity, and module origin. Database records are append-only—no edits, no soft-deletes.

Verify Audit Protocol →

Built and maintained by Shubham Kumawat

Independent engineering project. Direct line for security reviews, technical architecture inquiries, or procurement audits.

security@omnisentient.ai GitHub Repo

See the architecture in code.

Explore our system specifications, tamper-evident audit ledger protocols, and security controls.

View Architecture Specs → Read Trust Center