CortexOpen Source

Chat with your
running services.

An open-source AI agent that connects to your code, databases, logs, queues, and metrics. Deploy as a sidecar. Ask questions in plain English.

Get Started Request a Demo

Built-In Tools

10 tools. One conversation. Full picture.

Code Search

Natural language search across your entire codebase with TF-IDF ranking.

Health Checks

Real-time endpoint monitoring with SSRF protection and status aggregation.

Database Queries

Ask questions in English — Cortex generates read-only SQL with safety limits.

Log Analysis

Azure Monitor & AWS CloudWatch log search and pattern correlation.

Queue Inspection

Azure Storage Queues & AWS SQS — inspect messages, depth, and dead letters.

CI/CD Pipelines

GitHub Actions & Azure DevOps pipeline status, failures, and build history.

Infrastructure

Azure ARM & AWS CloudFormation resource listing and dependency mapping.

Metrics & Alerts

Azure Monitor & Prometheus queries with time-range filtering.

Incident Mode

Parallel investigation across all tools — timeline reconstruction in seconds.

Knowledge Base

Index runbooks, ADRs, and docs — searchable via natural language.

Architecture

Lightweight. Powerful.

On-Demand Retrieval

System prompt is ~200 tokens. Code is fetched only when needed — 80% token reduction vs. full-context.

Secure by Design

API key auth, read-only database enforcement, SSRF protection, SQL safety limits, non-root container.

Self-Hosted Sidecar

Deploy alongside each service as a Docker container, Kubernetes sidecar, or Cloud Run service.

LLM-Agnostic

Tool-Calling Agent

LLM orchestrates which tools to invoke. No hard-coded workflows — the model decides what to inspect.

Zero Idle Cost

~$0.001 per message at gpt-4o-mini rates. Scale to zero when inactive — pay only for what you use.

How it works

Deploy alongside your service. Start asking.

Cortex runs as a sidecar next to your service. Register your endpoints, databases, and log sources in a single YAML file. The LLM decides which tools to call — no hard-coded workflows.

Index your codebase in < 30 seconds

One YAML config — all secrets via env vars

Docker, Kubernetes, or Cloud Run

Apache 2.0 — free forever

cortex — chat

you: Why is the /api/users endpoint slow?

cortex: Let me check...

→ search_code("api/users endpoint")

→ query_database("slow query analysis")

→ search_logs("GET /api/users latency")

cortex: Found it. The user query in

routes/users.py:42 does a full table

scan — missing index on `created_at`.

P95 latency: 2.3s. Add an index and

it should drop to ~50ms. ✓

Coming Soon

Cortex Enterprise. For the fleet.

Federated multi-service intelligence with enterprise-grade access control.

Federated Hub

RBAC & SSO

Azure AD, Okta, Google OAuth. Team-based access control for every tool.

Audit & Compliance

Full audit logging, usage tracking, and token cost allocation per team.

Blast Radius Analysis

Cross-service dependency graph — know exactly what breaks before it breaks.

Try Cortex today.