Kyntra

AI-native full-stack engineering intelligence platform. Fully open-source.

Kyntra unifies API observability, frontend behavior analysis, automated test generation, and AI-driven code review into a single autonomous system that continuously observes, tests, and explains your software.

Self-host it with Docker Compose, Helm, or raw Kubernetes manifests. See docs/DEPLOYMENT.md for the full guide.

---

Architecture

kyntra/
├── apps/
│   └── dashboard/          # Next.js 14 — main UI
├── packages/
│   ├── shared/             # TypeScript types + utilities
│   ├── sdk/                # Node.js + Browser instrumentation SDK
│   ├── playwright-agent/   # Playwright E2E test generation
│   └── mcp-server/         # MCP server (Claude Code / AI assistant integration)
├── services/
│   ├── collector/          # Fastify — telemetry ingestion (port 3001)
│   ├── analyzer/           # Fastify + Claude AI — analysis engine (port 3002)
│   └── gateway/            # Fastify — unified REST API (port 3000)
├── infrastructure/
│   ├── helm/               # Helm chart for Kubernetes
│   ├── k8s/                # Raw Kubernetes manifests
│   └── load-testing/       # k6 load test (50K events/min)
└── docs/
    ├── PRD.md              # Full Technical PRD
    └── DEPLOYMENT.md       # Self-hosted deployment guide

Quick Start

Prerequisites

• Node.js 20+
• pnpm 9+
• Docker + Docker Compose
• PostgreSQL 16 (or use Docker)

1. Clone and install

git clone <repo>
cd kyntra
pnpm install

2. Configure environment

cp .env.example .env
# Edit .env and add your ANTHROPIC_API_KEY

3. Start infrastructure

pnpm docker:up         # Start PostgreSQL + Redis
pnpm db:migrate        # Run Prisma migrations
pnpm db:seed           # Seed demo project

4. Start all services

pnpm dev               # Start all services via Turborepo

Services start at:

• Dashboard: http://localhost:3003
• Gateway API: http://localhost:3000
• Collector: http://localhost:3001
• Analyzer: http://localhost:3002

---

SDK Integration

Install in your Node.js application:

npm install @kyntra/sdk

import { Kyntra } from '@kyntra/sdk';

// Initialize once at app startup
Kyntra.init({
  projectId: 'proj_xxx',
  apiKey: 'kyn_live_xxx',       // From Settings page
  environment: 'production',
  serviceName: 'my-api',
});

// All HTTP calls are automatically captured
// No other code changes required

Express middleware (inbound tracing)

import { kyntraMiddleware } from '@kyntra/sdk';
app.use(kyntraMiddleware());

Browser SDK

import { KyntraBrowser } from '@kyntra/sdk/browser';

KyntraBrowser.init({
  projectId: 'proj_xxx',
  apiKey: 'kyn_live_xxx',
});

---

Core Features

API Intelligence

• Auto-captures all HTTP requests/responses
• Normalizes URLs (/users/123 → /users/:id)
• Tracks latency (P50/P95/P99), error rates, health scores
• Auto-generates OpenAPI specifications from live traffic

AI Analysis (Claude claude-sonnet-4-6)

• Health Analysis: Score endpoints 0–100, detect degradation
• Anomaly Detection: Identify latency spikes and error surges
• Root Cause Analysis: Correlate signals across services
• Code Review: Risk scoring, security/bug/perf comments on PRs
• Test Generation: Complete, runnable test suites from traces

Frontend Agent

• Captures clicks, navigation, form submissions, errors
• Web Vitals (LCP, FID, CLS, TTFB)
• Fetch/XHR interception for API correlation

GitHub / Azure DevOps / Bitbucket Integration

• Webhook handler for pull request events across all three providers
• Auto-triggers AI code review on PR open/update
• Posts structured review comments + status checks back to the PR
• Zero-config: configure per-project via project settings JSON

---

API Reference

All endpoints are served by the gateway at http://localhost:3000.

Method	Path	Description
GET	/v1/projects	List projects
POST	/v1/projects	Create project
GET	/v1/apis?projectId=	API endpoint catalog
GET	/v1/traces?projectId=	Paginated trace history
GET	/v1/apis/:id/stats	Endpoint time-series stats
POST	/v1/analyze	Trigger AI analysis
GET	/v1/analyze/:id	Poll analysis result
GET	/v1/reviews?projectId=	Code review list
POST	/v1/reviews	Trigger code review
GET	/v1/tests?projectId=	Test case list
POST	/v1/tests/generate	Generate tests for endpoint
POST	/v1/reports/generate	Generate system health report
POST	/v1/github/webhook	GitHub webhook receiver
POST	/v1/azure-devops/webhook	Azure DevOps service hook receiver
POST	/v1/bitbucket/webhook	Bitbucket webhook receiver
GET	/v1/teams	List teams
POST	/v1/teams	Create team
POST	/v1/teams/:id/members	Invite member
POST	/v1/teams/:id/api-keys	Create API key
GET	/v1/alerts/rules?projectId=	Alert rules
POST	/v1/alerts/rules	Create alert rule
GET	/v1/alerts/events?projectId=	Alert event history
POST	/v1/alerts/events/:id/resolve	Resolve alert
GET	/v1/openapi?projectId=	Export OpenAPI 3.0 spec
GET	/v1/sla/current?projectId=	Current SLA status
POST	/v1/sla/compute	Compute SLA for period

---

Development

pnpm build             # Build all packages
pnpm typecheck         # TypeScript check all
pnpm test              # Run all tests
pnpm lint              # Lint all
pnpm db:studio         # Prisma Studio (DB browser)

---

Tech Stack

Layer	Technology
Monorepo	Turborepo + pnpm workspaces
Language	TypeScript 5.x (strict)
Backend	Fastify v4
Frontend	Next.js 14 (App Router)
Styling	Tailwind CSS + Recharts
Database	PostgreSQL 16 + Prisma
Cache	Redis 7
AI	Anthropic Claude claude-sonnet-4-6
Browser testing	Playwright
Container	Docker Compose

---

Roadmap

• M1 — SDK, Collector, Gateway, Dashboard skeleton
• M1 — Shared types, AI analyzer architecture
• M2 — Live AI health analysis + anomaly detection + BullMQ worker + trace viewer
• M3 — GitHub App + PR review posting + AI code review engine
• M4 — Playwright integration + UI test generation + frontend↔backend session correlation
• M5 — Multi-tenancy (Teams/RBAC), alert rules engine, OpenAPI export, Kubernetes manifests
• M6 — Kyntra MCP server (8 tools), Slack/PagerDuty notifications, SLA tracking

See docs/PRD.md for the full Technical PRD.

---

Self-Hosted Deployment

Kyntra is designed for easy self-hosting. See docs/DEPLOYMENT.md for the complete guide.

Docker Compose (quickest)

cp .env.example .env   # Set ANTHROPIC_API_KEY + JWT_SECRET
docker compose up -d
docker compose exec collector pnpm prisma migrate deploy

Helm (Kubernetes)

helm repo add bitnami https://charts.bitnami.com/bitnami
helm dependency update ./infrastructure/helm
helm install kyntra ./infrastructure/helm \
  --namespace kyntra --create-namespace \
  --set secrets.anthropicApiKey=sk-ant-xxx \
  --set secrets.jwtSecret=$(openssl rand -hex 32) \
  --set ingress.enabled=true \
  --set "ingress.hosts[0].host=kyntra.example.com"

Load Testing

# Verify 50K+ events/min throughput
brew install k6
k6 run infrastructure/load-testing/k6-scenario.js

---

Notifications

Kyntra supports alert notifications to:

• Slack — incoming webhook
• Microsoft Teams — incoming webhook with Adaptive Cards
• PagerDuty — Events API v2
• Generic webhook — HMAC-signed HTTP POST

Configure channels per alert rule in the Alerts UI.

---

License

Kyntra is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0).

• Free to self-host, modify, and use internally
• If you run a modified version as a hosted service, you must open-source your changes under AGPL-3.0
• For commercial/proprietary embedding, contact hello@kyntra.io for a commercial license

See LICENSE and CONTRIBUTING.md for details.

---

Built with the Kyntra master system — ApiMesh + Frontend Agent + AgnusAI.