# Cortex: Beyond a general AI workspace

Your models, your data, your machine. Use it as is, or we build it around your business.

## Hand it the task. Check the work.

### Your own agent, on your Mac and your phone

Cola runs your routines and long-running goals on your Mac and reports back. Message it from your iPhone and it keeps working while you are away.

- Message it from your iPhone
- Routines and goals on a schedule
- Reports when something changes

### Docs, slides and dashboards that stay editable

Ask for a report, a deck or a dashboard. The Agent writes it, you keep editing the same document, and any change it makes can be undone.

- Dashboards built from your files
- Export to Word, PowerPoint or PDF
- Undo any Agent edit

### Describe it, and the Agent designs it

Posters, social posts and pages appear on the canvas as real layers you can move and edit. Images and video are generated in place.

- Editable layers, not flat images
- Image and video generation
- Export to PNG, SVG or Figma

### Record a meeting, get the summary and to-dos

Both sides are recorded and transcribed on your Mac, speakers are told apart, and the summary links back to what was said.

- Transcribed on device
- Speakers told apart
- Decisions and to-dos with owners

### Answers from your own documents, with sources

Add policies, contracts and manuals. Every answer cites the page it came from, and the citations are checked before you see them.

- Cites the exact page
- Citations checked before you see them
- Your files stay in your library

## Bring your models. Pay for the right one.

Use your own API keys. this-that, the decision model from FLock.io, makes each small call in about 400 ms, so you spend less and wait less.

Works with: Claude, GPT, Gemini, DeepSeek, Kimi, Qwen, GLM, MiniMax, Seedance, and any OpenAI-compatible endpoint.

## It reaches the rest of your work

### Computer use

Clicks through your apps, asks before it sends.

### Routines

Scheduled tasks and long-running goals.

### Skills and connectors

Feishu, ERP, CRM and more over MCP.

Works with the tools you already use

- **Chat and bots:** Feishu / Lark, DingTalk, WeCom, WeChat, Telegram, Zendesk, Intercom, Discord, WhatsApp, LINE
- **Meetings:** Tencent Meeting, Granola, Zoom, Google Meet, Webex
- **Docs and knowledge:** Notion, Tencent Docs, Yuque, Google Drive, Jira and Confluence, Dropbox, Box, Airtable, Google Sheets, Obsidian, Coda
- **Projects and tasks:** Linear, Asana, Trello, ClickUp, Basecamp, Todoist
- **Mail and calendar:** Gmail, Google Calendar, Calendly, Mailchimp
- **Design and 3D:** Figma, Modao, Hyper3D Rodin, Tripo, Miro, Sketch, Framer, Webflow, Blender
- **Build and data:** Sentry, Supabase, Vercel, Cloudflare, Hugging Face, GitHub, GitLab, PostgreSQL, MongoDB, Snowflake, Zapier
- **Business and maps:** Stripe, Wix, Amap, Baidu Maps, Tencent Maps, Brave Search, Shopify, HubSpot, PayPal, Square, Xero, QuickBooks
Plus any MCP server you add.

## From foundation to outcome

The Cortex you download is the foundation. We add the scenarios your business runs on, and what comes out is material you can check and reuse.

- **03 Output — Outcomes:** Contract review report, Due diligence brief, Operations dashboard
- **02 Scenarios — Your business:** Retail, Legal, Manufacturing, Healthcare
- **01 Foundation — Cortex:** Local knowledge, Tiered permissions, Privacy

## Built on Cortex

Five workflows built on the same foundation. Cortex does the routine steps, and a person signs off.

### Fashion retail

For merchandising and marketing teams: product shots, model shots, detail pages and campaign video for every new drop, in the brand’s own look.

- Brief and brand theme
- Product and model shots
- Detail page and video
- Team reviews
- Batch export

### Legal

For in-house and firm lawyers: pull out parties, amounts, term, payment and liability, flag risky clauses and draft the review memo. Files are redacted on your Mac first.

- Redact the contract
- Extract key clauses
- Flag risky clauses
- Draft review memo
- Lawyer signs off

### Federated learning

Agentic FL, built for FLock. Describe a federated training run in chat, and Cortex sets it up, runs it and reports back.

- Describe the run
- Configure nodes
- Owner approves
- Train
- Report results

### Manufacturing

For plant and supply-chain teams: answers from SOPs and equipment manuals, a daily production and quality dashboard, and alerts on late supplier deliveries.

- Load SOPs and manuals
- Answer with the page
- Daily quality dashboard
- Flag late deliveries
- Supervisor signs off

### Healthcare

For clinics and research teams: patient records redacted on your Mac, assessment notes summarized, and draft reports and training plans for clinicians to approve.

- Redact records
- Summarize assessments
- Draft report
- Doctor reviews
- Training plan

## Cortex, built around how your team works

We take one workflow your team runs every week and build Cortex around it, from the first conversation with your team to the hand-over.

- **01 Map the workflow:** We sit with the people doing the work and pick the steps worth handing to AI, and the ones a person should still check.
- **02 Build on your real cases:** We build against your own documents and past work, and test until the output is something your team would sign off.
- **03 Roll out with one team:** One team uses it for real work first. What they change goes back into the build before more teams join.
- **04 Hand it over:** Your team owns what we built and can keep editing it. We train the people who run it.

## Enterprise grade, on your infrastructure

What a custom build can include. Even without one, your files and chats stay on your own machines.

### Private deployment

Run Cortex and its models inside your own network. Files, chats and model calls stay on infrastructure you control.

- On-premises servers
- Private cloud (VPC)
- Offline installer, no internet needed

### Custom agent harness

We build the agent around your workflow on Cortex’s agent runtime: the tools it can call, the procedures it follows and where it must stop for a person.

- Tools and MCP connectors for internal systems
- Skills for your procedures
- Approval steps and sub-agents

### Proprietary models

We fine-tune open-weight models such as Qwen, DeepSeek and GLM on your data, so the model knows your terms, formats and decisions.

- Trained on your data, run on your hardware
- Federated training with FLock when data can’t move
- Small, fast models built for one task

### Enterprise permissions

Access is designed around your org chart: who can use which knowledge, which tools and which actions.

- Three tiers, from staff to the owner’s cockpit
- Knowledge scoped by team and project
- Confirmation before risky or outbound actions

## Download Cortex

### Mac, Apple silicon

Latest release

Free to download. You pay your model provider directly. Requires macOS 13 or later. Intel Macs are not supported yet.

### Windows, Linux and mobile

Coming soon.

- Windows
- Linux
- iPhone
- Android

## Questions

Anything else? Talk to us.

**What does Cortex cost?**
Cortex is free to download. You bring your own API keys and pay your model provider directly, or run a model on your own machine through any OpenAI-compatible endpoint. Custom builds for teams are quoted per project.

**Where does my data go?**
Your files, chats, knowledge base and meetings stay in a folder on your Mac. A request leaves your Mac only when it goes to the model provider you chose. Meeting transcription and the redaction app can run entirely on device.

**Which models can I use?**
Claude, GPT, Gemini, DeepSeek, Kimi, Qwen, GLM, MiniMax and any OpenAI-compatible endpoint. In auto mode, this-that picks a model for each task from the ones you set up.

**What can the agent do on my computer?**
It reads the folders you give it and runs commands in a macOS sandbox. Risky steps, like deleting files or sending messages, wait for you to confirm. Plan mode lets it read and plan without changing anything.

**Does it work on Windows, Linux or my phone?**
Cortex runs on Macs with Apple silicon today. Windows, Linux and mobile apps are coming soon.

**How is it different from ChatGPT?**
Cortex works on your own files on your Mac and keeps what it makes as documents, decks and dashboards you can keep editing. It runs routines on a schedule, records meetings, and connects to the tools your team already uses.

## Links

- Website: https://cortex.flock.io
- Download: https://github.com/FLock-io/cortex-releases/releases/latest
- Company: https://www.flock.io
- Contact: hello@flock.io
- Privacy: https://cortex.flock.io/privacy.html
- Terms: https://cortex.flock.io/terms.html
