CORTEX — OPEN SOURCE BYOK WORKFLOW

Your AI writes code.
Cortex understands
the codebase.

Open-source codebase onboarding in minutes, grounded in your local dependency graph. Bring your own AI provider only if you want semantic analysis or onboarding-doc generation.

View Sample Output Read Quick Start

Open source  ·  BYOK  ·  No SYKE-managed model billing

Onboarding usually burns 20–40
senior-engineering hours per hire.

Every new developer — and every new AI agent — spends days or weeks learning which files do what. That knowledge usually lives in senior engineers' heads until someone writes it down.

WITHOUT SCAN_PROJECT
WEEK 1–2 New dev reads random files, asks seniors where everything lives.
WEEK 3 Still doesn't know which files are dangerous to touch alone.
WEEK 4 First confident PR — maybe. Seniors spent 20+ hours explaining architecture.
Cost per new hire
$4,000–$8,000
in lost productivity + senior time
WITH SCAN_PROJECT
MINUTE 0 Run scan_project. One command.
MINUTE 5 Full architecture doc generated — module map, danger zones, entry points, circular dependencies.
DAY 1 New dev ships their first PR with confidence. Already knows which files to never touch alone.
Incremental cost
Your provider usage
depends on model choice, prompt size, and provider limits

10-section onboarding document.
Generated in minutes.

scan_project analyzes your full dependency graph and feeds structured data to your chosen AI model. The output is a comprehensive markdown document, ready to share with the entire team.

Project Overview Purpose, stack, team conventions, and what the project actually does.
Architecture Overview High-level system design, layer boundaries, and data flow.
Repository Structure Directory map with plain-English explanations of every folder.
Module / Component Map Which modules own what, their public APIs, and how they interconnect.
Key Concepts & Domain Model Core business logic, domain terminology, and the mental model to build from.
Development Workflow How to run, test, build, and deploy the project locally and in CI.
Architectural Decisions Why the codebase is structured this way — context that rarely exists in docs.
Cross-Cutting Concerns Auth, logging, error handling, and other patterns that cut across modules.
Danger Zones & Common Tasks High-risk files, known gotchas, and step-by-step guides for frequent changes.
Quick Reference Commands, key files, important contacts, and everything to bookmark on day one.

Three steps. Five minutes.

Cortex uses BYOK — your own API key, your own AI provider. No data sent to SYKE servers, and no SYKE-managed markup on model usage.

Add your AI key

Set one environment variable in your project or shell. Cortex works with any of the major providers — choose whichever provider already fits your workflow.

GEMINI_API_KEY OPENAI_API_KEY ANTHROPIC_API_KEY

Your key, your provider, your billing relationship. Exact cost depends on model choice and prompt size.

Run scan_project

SYKE traverses your entire dependency graph — every file, every import, every module relationship. That structural data is fed to your AI model with a purpose-built prompt. 

// In Claude Code, Cursor, or any MCP client:
scan_project({ output: "ONBOARDING.md" })

Get your onboarding doc

A comprehensive markdown document drops in your project root. Share it with your new hire, your AI agent, or your future self. Regenerate it any time the architecture changes.

✓ ONBOARDING.md — ready to commit, share, or paste into your AI chat.

Turn repeat onboarding work into a reusable document.

The win is time and repeatability: generate context once, refresh it whenever the architecture changes, and stop explaining the same map by hand.

ROI_CALCULATOR — PER NEW HIRE
Senior engineer walkthrough time 20–40 hours
Time to regenerate the document after major changes Minutes
Manual knowledge transfer risk Stale, inconsistent, tribal
BYOK setup overhead One provider key
Repeatability Re-run any time
Hours reclaimed per onboarding cycle 19.5–39+

"The core server stays free and local. BYOK only changes who handles optional model usage."

ai_analyze — Your AI Safety Net

Before modifying any file, get a semantic analysis that understands not just the file — but every file that depends on it.

ai_analyze

Reads the full source of a file and all of its direct dependents, then asks your AI model to identify what might break, why, and how to safely make the change.

Flags broken imports and type mismatches before you run a single build.

Identifies interface violations caused by upstream changes.

Explains ripple effects in plain English — not just file names.

Works with Gemini, OpenAI, or Claude. BYOK, always.

"Think of it as a senior code reviewer that never sleeps, never skips context, and never gets annoyed at your questions."

Open source locally.
Bring AI only when needed.

One command. One document. Onboard any developer — or AI agent — to your entire codebase with a local-first, BYOK workflow.

View Sample Output Open Repository

Open source  ·  BYOK  ·  Run locally

See a sample onboarding document →