Skip to content
| Marketplace
Sign in
Visual Studio Code>Programming Languages>Cora: AI Secured, On-device Agentic Coding ToolNew to Visual Studio Code? Get it now.
Cora: AI Secured, On-device Agentic Coding Tool

Cora: AI Secured, On-device Agentic Coding Tool

CodeMate

|
671 installs
| (14) | Free
AI coding agent that generates code from natural language, automates tasks, and runs terminal commands. Features inline autocomplete, browser automation, automated refactoring, and custom modes for planning, coding, and debugging. Uses a local OpenAiCompatible model endpoint.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

CORA

Autonomous AI coding agent that plans, architects, and builds production-ready software inside VS Code.

CORA is not a code completion tool. It is an agentic coding assistant that understands what needs to be built before writing a single line mapping requirements, making architectural decisions, resolving dependencies, and executing the build in the correct order. Every step is visible. Every change requires your approval.

Install from VS Code Marketplace · Website · Twitter · YouTube


What CORA Does

Most AI coding tools help you write faster. CORA handles everything that happens before typing: understanding the real requirements behind a prompt, surfacing what was not asked for but will eventually be needed, choosing the right stack with explicit reasoning, and building in dependency order rather than all at once.

Given a prompt like:

Build a SaaS analytics dashboard with multi-tenant support and real-time charts

CORA in action

CORA does not start generating files. It first maps the full scope identifying implicit requirements like invitation flows, API key management for machine-to-machine auth, and tenant-scoped queries on every database call. It then produces an architecture plan you can review and push back on. Once confirmed, it builds layer by layer: foundation, services, data layer, frontend, integration, validating each layer before proceeding to the next.

The result is a complete, architecturally sound codebase. Not a scaffold with placeholders. Not disconnected snippets. Not a project that still needs days of wiring.


Why Developers Choose CORA

Basic AI Coding Tools CORA
Scope File-level suggestions Full project-wide planning and execution
Planning None Architecture docs, stack decisions, dependency-ordered task graphs
Transparency Generates without explaining Every decision explained before action
Control Accept or reject completions Approve every file change, rollback anytime
Context Open file only Full codebase, Git history, MCP-connected tools
Configuration None Skills, rules, and workflows scoped to your project

Modes

CORA operates in five modes. Select the mode before sending a message, or use Auto and let CORA decide.

Modes in CORA

Auto

Orchestrator mode. CORA selects the right mode Plan, Ask, Code, or Review based on what the prompt requires, and switches mid-task when needed. Use when you want end-to-end progress without micromanaging every step.

Orchestrator in CORA

Plan

Analyzes your codebase and produces architecture documents, stack decisions, and a dependency-ordered task breakdown. No code is written. Use when designing a new feature, planning a refactor, or evaluating technical approaches before committing.

Code

Writes, modifies, and refactors code with full codebase context. The primary execution mode. Use when building features, fixing bugs, or making targeted changes across multiple files.

Ask

Answers questions about your codebase, explains past decisions, and provides guidance. Makes no changes. Use when you need to understand the codebase, explore options, or get a second opinion before acting.

Review

Audits recent changes for correctness, consistency with existing patterns, performance, and security. Use before committing, merging, or deploying.


Configuration

CORA's behavior is shaped by configuration files stored in your project version-controlled, team-shareable, and scoped per repository.

Settings in CORA

Skills

Markdown files that give CORA domain knowledge specific to your project: API design patterns, component conventions, testing standards, naming rules. Define once, apply across all sessions.

Skills in CORA

.cora/skills/api-design.md .cora/skills/testing-standards.md .cora/skills/security-review.md

Rules

Hard constraints applied to every interaction: forbidden patterns, required file structures, architectural boundaries. CORA will not violate rules.

Rules in CORA

Workflows

Step-by-step processes for repeatable tasks, invoked via slash commands.

Workflows in CORA

.cora/workflows/scaffold-feature.md .cora/workflows/pr-prep.md .cora/workflows/run-migration.md

MCP Servers

JSON configuration to connect CORA to external tools GitHub, Slack, databases, internal APIs, CI pipelines. Supports per-project setup and secure authentication.

MCP Servers in CORA - Setup MCP Servers in CORA - Configuration


Core Capabilities

Full codebase awareness CORA reads your entire project, not just the open file. It understands module boundaries, existing patterns, import graphs, and Git history before making any suggestion.

Structured task planning Complex prompts produce a dependency-ordered task graph before any code is written. You can review, reorder, and modify the plan before execution begins.

Architecture diagrams CORA generates Mermaid diagrams for system architecture, data flows, and component relationships during planning sessions.

Change approval and rollback Every file modification is presented as a diff. Nothing is applied without explicit approval. Checkpoints are maintained so you can roll back to any prior state.

50+ model support Works with Claude, GPT, Gemini, DeepSeek, and other leading models. Switch models per session or per mode.

Local-first and privacy-aware Designed for teams with data residency requirements. Supports self-hosting and private model endpoints.

Commit messages Generates commit messages from staged diffs. Understands intent and context, supports conventional commit format, and integrates with VS Code's built-in Git tooling.


Getting Started

Install

Open the Extensions panel in VS Code (Cmd+Shift+X on Mac, Ctrl+Shift+X on Windows/Linux), search for Cora Agent, and install. Or click Install at the top of this page.

Sign in with your CodeMate account when prompted. The CORA panel will appear in the activity bar.

Run Your First Build

  1. Open a project folder in VS Code, or create an empty folder for a new project.
  2. Open the CORA panel and select Auto mode.
  3. Write a prompt describing what you want to build. Include the domain, key features, and any hard technical constraints.
  4. CORA will output an architecture plan before writing any files. Review the stack choices and surfaced requirements. Ask questions or push back this is the right time.
  5. Confirm the plan. CORA builds layer by layer, writing files into your workspace as each layer is validated.

Continue the conversation after the initial build to add features, ask questions about the generated code, or run a Review pass before merging.


Supported Languages and Frameworks

JavaScript, TypeScript, Python, Go, Java, Kotlin, Rust, C, C++, C#, Swift, PHP, Ruby, Dart, SQL, Shell, R, Julia, React, Next.js, Vue, Svelte, Angular, Tailwind, Node.js, FastAPI, Django, Spring, Docker, Terraform, YAML, and more.


Links

  • Website
  • Twitter
  • YouTube
  • Email
  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2026 Microsoft