UX PROTOTYPE v2. Not the real Cortex. Nothing here is saved or sent anywhere.
Home
CORTEX
One source of truth for how work gets done.
Your agents can change. Your models can change. Your harnesses can change. Cortex stays. Rules, skills, workflows, and standards live here once, and every AI you connect gets the same truth.
Cortex does not remember for your AI. Your conversations and personalization stay with each AI. Cortex tells your AI how the world works.
About 5 minutes. Skip anything. Nothing leaves this demo.
Bring me what you have already built.
Your project knowledge is scattered across files that were never designed to be a system. Tap each source you want Cortex to absorb.
Cortex is reading your project.
Reading 5 sources...
Identifying rules, skills, workflows, and conventions. Checking for conflicts. Flagging what is outdated.
Here is what Cortex understood.
Review each card. Nothing becomes authoritative until you say so.
One conflict found. Two sources disagree on database naming. Cortex suggests the newer one. You decide.
THE PAYOFF
Five versions of the truth become one.
Before
CLAUDE.md
6 Cursor rule files
README
11 docs
5 skills and workflows
Scattered. Some conflicting. One file still says FTP.
After
21
items in one authoritative system
5 rules, 3 skills, 2 workflows, 6 knowledge items, 5 permissions. Every connected agent reads from here.
Change a rule here once, and Cortex translates it into each agent's own format. You never maintain five versions again.
Connect Claude Code.
Cortex does not replace what Claude Code remembers. It gives Claude Code the project knowledge it needs, in Claude Code's own format.
1
Preview
2
Apply
3
Verify
# your existing CLAUDE.md stays exactly as it is<!-- CORTEX:START ecclesio-rules v1 --># Ecclesio rules, from Cortex- Database tables use snake_case.- Production deploys need your approval.- No em dashes in user-facing copy.<!-- CORTEX:END -->
# .muse/rules/cortex.mdc (Cortex-owned)---description: Ecclesio project rulesglobs: **/*---# Rules from Cortex- Database tables use snake_case.- TypeScript strict. No silent any.- Component files use PascalCase.
# ChatGPT project instructions (Cortex-synced)You are working on Ecclesio, a church operationssoftware project. Follow these project rules:- Database tables use snake_case.- Release process: PR, CI, staging, approval, production.- Giving data is sensitive. Never include it in output.
Advanced: adapter details
adapter: claude-code (Type A, file-based)
owns: marker blocks only. your edits outside blocks are never modified.
sync: session-start check + desktop watcher + scheduled reconcile
connection id: cx_conn_9f31ac (demo)
Writing marker block into CLAUDE.md...
✓Claude Code is connected
Current. It reads the Ecclesio source of truth at every session start.
SETUP COMPLETE
Ecclesio is ready.
Here is what you just built.
5
rules
3
skills
2
workflows
6
knowledge
5
permissions
1
agent connected
✓5 scattered sources absorbed into one authoritative system
✓Claude Code connected and current
Good morning, Jon.
Needs your decision
Needs your attention
Ecclesio: Cursor needs 3 updates
Database naming, release process, member definition
›
Meta Muse has not synced in 3 months
14 things changed since it last connected
›
What changed
Prototype demo: six months later
Half a year of operational drift, then reopen Claude Code and watch it become current. Try it →
Projects
Each project holds one authoritative system: its rules, skills, workflows, knowledge, and permissions. Connected agents read from it.
Agents
Every agent works from the same Cortex truth, translated into its own format. Cortex never touches what an agent remembers.
Connect a new AI and watch it become productive.
Connect Codex.
Codex has not worked on Ecclesio before. Here is what Cortex can provide, so it starts productive on day one.
From the Ecclesio source of truth
✓Project knowledge (6 items: what Ecclesio is, sensitive data rules, member definition...)