Build software
with an AI agent

that understands your project.

Cortex AI IDE is a cross-platform agentic coding workspace. Describe what you want, and the agent plans the change, edits across files, runs terminal commands, and verifies the result. Each change lands as a reviewable diff.

Download → Source on GitHubGitHub

In short

A free, cross-platform AI coding agent. It reads your project, edits multiple files, runs and verifies commands, and shows each change as a reviewable diff. Describe a feature and it ships the code.

Who it's for

For developers and small teams building real software: SaaS, e-commerce, fintech, internal tools, and open source. Use it when you need an agent for multi-file edits, terminal work, and verified results rather than autocomplete. No subscription required; code stays on your machine.

What you can build

Python and Django web apps, native desktop and Android apps, and Electron tools in TypeScript, Rust, Go, or your stack. From SaaS platforms and REST APIs to CLIs, automation scripts, and AI-powered features.

Animated preview of the Cortex AI IDE and its agent terminal working on a task, over a painted landscape background.

Cortex AI IDE

Git Review (10)

main

7 unstaged 2 untracked

Msettings.py+1-1

Murls.py+8-1

Mstyles.css+489-62

Mmain.js+140

Mindex.html+139-16

Mviews.py+7

Murls.py+2

Udemo.html

Ubackground.png

Build 3.0.13 release page

Readchangelog.md

Readcortex/views.py

Thought6s

I’ll build a serif release page that matches the site’s voice — entries newest first, badges inline.

templates/cortex/release.html+52−0

cortex/urls.py+6−0

Questions‹›

How should the release entries be ordered?

1Newest first, badges inline

2Grouped by feature area

3One paragraph per version

Skip

Plan, search, build anything…

Agent MiMo V2.5 Pro ↑

requirements.txt× runtime.txt× admin.py× apps.py× urls.py× views.py

C:▸ Users▸ dev▸ projects▸ cortex-ide▸ cortex Copy Path

1"""

2cortex/views.py — Public Site Views

3=============================================

5View functions that render the public pages of the Cortex

6website: landing, features, changelog, and download.

7"""

9from django.http import Http404, JsonResponse

10from django.views.decorators.http import require_GET

11from django.core.paginator import Paginator, EmptyPage,

12PageNotAnInteger

13from django.db.models import Count, Sum, Q

14from django.shortcuts import render, get_object_or_404, redirect

15from django.utils import timezone

16from django.contrib import messages

18from api.models import Subscription, UsageLog, Payment,

19Release, DownloadEvent

Powershell 1▾ + New × Clear ⟳

Watching for file changes with StatReloader

System check identified no issues (0 silenced).

August 19, 2026 - 20:26:14

Django version 6.0.6, using settings 'config.settings'

Starting development server at http://127.0.0.1:8000/

Quit the server with CTRL-BREAK.

WARNING: This is a development server. Do not use it in a production setting. Use a production WSGI or ASGI server instead.

[2026-08-19 20:27:52,327] WARNING django.request Not Found: /.well-known/appspecific/com.chrome.devtools.json

cortex\views.py changed, reloading.

System check identified no issues (0 silenced).

Cortex Agent

~/projects/cortex-ide

Release 3.0.13 changelog

❯cortex agent "ship the 3.0.13 changelog entry"

⬢Read cortex/templates/cortex/changelog.html

⬢Thought 3s

⬢Searched for 'release notes format 3.0.x'

⬢Edit changelog.html +14 -1

cortex/templates/cortex/changelog.html +14 -1

→ Add a follow-upctrl+c to stop

MiMo V2.5 Pro
/ for commands · @ for files

13 Agent tools

8 BYOK providers

MCP Server support

Cross-platform Native apps

Overview

What is Cortex AI IDE?

The Agent Your Keys 250+ Skills Project Memory

Describe it. The agent ships it.

Cortex runs the full agentic loop — read, think, edit, verify. State your intent in plain language and the agent plans the change across files, runs terminal commands, and checks the result. Your codebase is the context and your keys stay local. Each change lands as a reviewable diff. See how the agentic loop works step by step.

Cross-platform local-first diff review before merge

Read full codebase indexed

Think plan before touching files

Edit multi-file, terminal included

Verify runs it, checks the result

Your keys, your models, your budget

Cortex is not tied to a single model provider. Connect your own API keys and switch models when your task, budget, or preferred provider changes. You pay the model provider directly. Cortex does not mark up or resell your BYOK usage. New to the model? Read what BYOK means and why it changes the maths.

0% markup keys stay local swap anytime

openai-compatible

token plans + direct apis

free models via OpenRouter

MiMo DeepSeek Claude GPT-5 OpenRouter Qwen

250+ skills

More than a general-purpose agent

Cortex includes 250+ skills for focused development workflows such as debugging, refactoring, documentation, SEO, design reviews, and project-specific tasks.

Enable from Settings

Toggle skills on from the Settings → Skills panel. Enabled skills guide the agent on every message.

Call with / commands

Type / in the chat input to call a skill by name (one or several) and watch the active skill in the status bar.

Customize or add your own

Edit any skill in the built-in Monaco editor, or add compatible skill files to your own projects.

Project memory

Context that stays with your project

Local, per-project context

Cortex keeps project context locally so you can return to work without rebuilding the same context from scratch. Project memory is stored under .cortex/memory/, while chat history is kept locally in a project-specific SQLite database. Everything stays separate, local, and available when you reopen the IDE.

.cortex/memory/

project.sqlite

restored on reopen

Xiaomi MiMo V2.5 DeepSeek V4 Anthropic Claude OpenAI GPT-5 OpenRouter Alibaba Qwen Google Gemini InferenceHub Xiaomi MiMo V2.5 DeepSeek V4 Anthropic Claude OpenAI GPT-5 OpenRouter Alibaba Qwen Google Gemini InferenceHub Xiaomi MiMo V2.5 DeepSeek V4 Anthropic Claude OpenAI GPT-5 OpenRouter Alibaba Qwen Google Gemini InferenceHub

Cortex AI IDE editor with agent panel, split workspace

Cortex AI IDE agent chat with thought stream and tool calls

From idea to working code

Cortex is built for more than autocomplete. Give the agent a goal and it can explore your project, understand the surrounding code, edit multiple files, run commands, inspect the output, and iterate when something fails. You can review changes before they are finalized, or use Plan Mode to see the implementation approach before editing begins.

How agents work →

Polish the pricing page hero

Thought6s Tighten the headline and lift the CTA contrast.

Read pricing.html · styles.css

Searched 4 files for “hero-cta”

pricing.html+38−12

styles.css+21−6

Questions Keep the gold CTA, or try azure? GoldAzure

Ask a follow-up…

pricing.htmlstyles.cssplan.md

cortex / templates / pricing.html

Pricing hero polish 1. Headline: one line, no wrap at 1280px. 2. CTA: gold fill, azure on hover. 3. Verify: screenshot + contrast check.

Tasks 0/3

Tighten headline Lift CTA contrast

Verify in preview…

One workspace for real development

Everything you need to build

Work across the full codebase, run and verify commands, and use the tools you already know without leaving the editor.

Tools Codebase Run & Verify MCP Native Security

Built-in developer tools

Use file read/write/edit, terminal commands, grep, glob, semantic search, web search, Git, LSP tools, task tracking, and more from the agent workflow.

agent · toolchain executing edit

readwriteeditbashgrepglobsearchwebgittodolooplsp+3

❯ edit views.py +14 −1

Work across the full codebase

Semantic search helps Cortex find code by meaning, not just exact text. The agent follows relationships across files and works with the architecture of your project.

Run and verify

Cortex can execute commands, read terminal output, run tests, and use the results to correct its work.

powershell

❯pytest tests/ -q

12 passed in 3.4s

✓verify gate passed

MCP support

Connect Model Context Protocol servers to extend Cortex with your own tools and services.

Native across platforms

Cortex runs directly on your operating system. No WSL or container is required.

Microsoft Store –□✕

C Cortex AI IDEProductivity · Free Get

installing… 62% no WSL · no container

Security

Your keys stay on your machine.

API keys are stored in your operating system's secure credential store and protected with native OS encryption. BYOK requests go directly from your machine to the selected AI provider instead of being proxied through Cortex servers.

✓Local key storage. Keys are stored in your operating system's secure credential store.
✓Direct provider requests. Your model traffic goes directly to the provider you choose.
✓No plaintext key files. Keys are not written to ordinary configuration files or logs.
✓No provider lock-in. Change providers or models whenever you want.

cortex · key vault sealing DeepSeek

❯seal key → DPAPI + AES-256-GCM

✓stored in your OS secure credential store

plaintext on disk: 0 bytes · sent to Cortex: 0 bytes

✕no proxy · no middleman · 0 bytes via Cortex

❯ request → api.deepseek.com

Pricing

Is Cortex AI IDE free to use?

What you get with and without a subscription

Every core coding feature is free with your own API key. The subscription adds managed Cortex services on top.

Feature	Free (BYOK)	Optional Subscription
AI coding agent
Full codebase context
Terminal & Git tools
Hosted semantic search
Web search & fetch (RAG)
Voice input & OCR
Priority support (24h)

Build with an agent
on your side

Open your project, choose your model, and give Cortex a goal. Review the plan, inspect the diff, and ship when you are ready.

Download View Plans

Cross-platform · 64-bit · BYOK, 8 providers

TypeScript JavaScript Python Go C# C++ Java Rust HTML CSS Markdown Spreadsheet PDF TypeScript JavaScript Python Go C# C++ Java Rust HTML CSS Markdown Spreadsheet PDF TypeScript JavaScript Python Go C# C++ Java Rust HTML CSS Markdown Spreadsheet PDF

What do developers ask about Cortex?

The most common questions cover BYOK setup, supported providers, model switching, data privacy, and what you can build. Each answer below is a straight, no-fluff response.

Get answers to commonly asked questions

Straight answers for developers who build with Cortex every day.

BYOK means Bring Your Own Key. You connect an API key from a supported AI provider and pay that provider directly for model usage.

Cortex supports eight provider integrations: Xiaomi MiMo, DeepSeek, Anthropic, OpenAI, OpenRouter, Alibaba Qwen, Google Gemini, and InferenceHub. Available models and context limits depend on the provider and model you select.

Yes. Enable the providers you want in Settings, then choose between available models from the agent interface.

No. BYOK keys are stored locally in your operating system's secure credential store and are not sent to Cortex servers.

For BYOK model requests, Cortex connects directly from your machine to the selected provider over TLS rather than routing the request through Cortex servers.

Open Settings → Models & Providers, select a provider, and enter your API key.

Cortex can work across many software projects, including Python and Django backends, Node.js services, web applications, Electron apps, React Native projects, Android apps, and desktop software. Its usefulness depends on the tools, language support, model, and project environment you choose.

Each project can keep its own local memory under .cortex/memory/. Chat history is also stored locally so context remains associated with the project.

Skills are reusable instruction sets for specialized workflows. Cortex includes 250+ built-in skills, and you can enable, inspect, customize, or add skills for your own projects.

The optional subscription provides managed Cortex services such as hosted semantic search and embeddings, web search and fetch, OCR fallback, MCP services, voice input, and priority support. The subscription is not required for BYOK coding.

Yes. The IDE and BYOK workflow can be used without a required subscription.

Cortex is available now for Windows and Linux: it runs natively on Windows 10 and 11 and Linux, with no WSL or container required, and your project stays on your machine. macOS is available via the official build guide.