16 AI agents.
Every PR. In parallel.
Not one generic AI doing a review. 16 specialized agents — each an expert — running simultaneously. Security, performance, architecture, FinOps, resilience, and more.
AI finds the issues. Human-defined guardrails decide what blocks. Connected to your CI/CD, incidents, and cloud costs.
Three-Layer Architecture
01
16 AI Agents
Each specialized in one domain. Running in parallel with structured findings, confidence scores, and inline diffs.
02
Policy Engine
Toggle-based guardrails. Block, warn, or require approval. No Rego. No YAML. Templates included.
03
Human Guardrails
AI finds issues. Humans decide what matters. Override flow with approval and full audit trail.
16 Specialized AI Agents
3-pass pipeline: regex/AST scan → LLM triage → deep analysis. Minimizes cost while maximizing coverage.
Security
HeavyOWASP Top 10, injection, auth
Performance
HeavyN+1 queries, memory leaks, blocking I/O
Architecture
HeavyCoupling, patterns, circular deps
Design Patterns
HeavySOLID, GoF patterns, code smells
Test Coverage
HeavyBranch coverage, test gaps, critical paths
AI Code Quality
HeavyHallucinated APIs, over-engineering
FinOps
MediumCloud cost impact, over-provisioning
API Design
MediumREST conventions, RFC compliance
Resilience
MediumCircuit breakers, graceful degradation
Code Quality
MediumMaintainability, complexity, duplication
Bug Detection
LightLogic errors, race conditions, null safety
Observability
LightLogging, metrics, tracing gaps
Checkstyle
LightNaming, formatting, conventions
Secrets
LightHardcoded credentials, API keys, tokens
Dependencies
LightCVEs, deprecated pkgs, license issues
Feedback
LightHistorical context, past incidents
PR Health Score
Every PR gets a score. 0–100.
Composite score based on severity-weighted findings across all 16 agents. Not a vague pass/fail — your team sees exactly where the risk is.
Score = 100 - Σ(severity × count)
Critical: -25 · High: -15 · Medium: -8 · Low: -3
Platform Capabilities
Guardrails
Toggle-based quality gates. Block, warn, or require approval. No code, no Rego. Templates: "No Critical Findings", "Security Required", "No Secrets".
CI/CD Diagnosis
AI root cause analysis of pipeline failures. Failed job breakdown with actionable code suggestions to fix builds.
RCA Hub
Root cause analysis for production incidents. Links PRs to affected services with investigation timelines.
FinOps
Cloud spend trends by service/team/environment. AI cost optimization suggestions. PR cost impact estimation.
40+ Integrations
AWS, GCP, Azure, Datadog, PagerDuty, Jira, Slack, Sentry, and more. OAuth, API Key, IAM Role auth.
Analytics
Review trends, finding severity over time, team performance, agent effectiveness, health score trends per repo.
RAG + pgvector
Vector embeddings of past findings. Context-aware reviews that learn from your codebase history and team conventions.
GitHub + GitLab Native
Auto webhook setup, inline PR comments, status checks, file-level annotations. Feels native.
.terraqube.yml
Per-repo config. Choose agents, set severity thresholds, exclude paths, customize behavior per repository.
Architecture — 6 Microservices on GCP
API Service
Python/FastAPIREST API, webhook receiver, billing (Stripe), team management
Review Orchestrator
Python/FastAPIPR events → diff fetch → agent pipeline → guardrail evaluation → post results
Agent Service
Python/FastAPI16 AI agents, 3-pass pipeline, structured findings with confidence scores
Pipeline Service
PythonCVE crawl (6h), package health (daily), threat intel (12h), RAG embeddings
Worker Service
PythonHealth watchdog, content engine, email lifecycle, cleanup, prompt tuning
Web Dashboard
Next.js 16 / React 19Reviews, repos, guardrails, analytics, integrations, billing, settings
Why nobody else does this.
vs Linters
Check syntax rules. Terraqube's 16 agents understand architecture, security, performance, resilience, and FinOps — things no rule-set can express.
vs GitHub Copilot
Helps write code. Terraqube reviews what's written with domain-specific agents that catch design problems, not just syntax errors.
vs SonarQube / CodeClimate
Count metrics. Terraqube reasons semantically AND connects to CI failures, incidents, and cloud costs.
vs Manual Code Review
Humans can't run 16 parallel analyses on every PR. Terraqube augments your team — consistently, on every commit.
Plans
Free
3 repos, basic agents
Pro
Unlimited repos, all agents, guardrails
Enterprise
SSO, audit logs, custom agents
Launching 2026. Get early access →