AI IDE

The most capable AI IDE.

A VS Code-based IDE built for serious work: five operating modes, 9+ LLM providers, autonomous background agents, Council mode for multi-agent deliberation, and text-to-application generation. Every AI edit staged for human review. Free on desktop, or cloud-hosted for teams.

Book a demo See pricing →

A serious tool for serious work.

FREE DESKTOP DOWNLOAD

No accounts, no signup, no strings. Download Calliope AI IDE and start working immediately — personal, professional, or commercial use, completely free.

Built on the open-source VS Code core, stripped of the surveillance. All telemetry disabled, no phone-home, no nagware. Runs entirely local with your API keys staying on your machine — use Ollama for fully air-gapped development. Your existing VS Code extensions still work.

The right AI mode for every task — and you always stay in control.

Switch among 9+ providers — OpenAI, Anthropic, Google Gemini, Groq, DeepSeek, Mistral, xAI Grok, Together AI, AWS Bedrock — plus Ollama for local models. Set different models for chat, autocomplete, and agent tasks. Bring your own keys.

Agent Mode — Autonomous code writing and file manipulation. Background agents work while you focus on architecture.
Ask Mode — Read-only exploration with web search. Investigate without changing code.
Plan Mode — Design-first with human approval gates. AI proposes, you approve before execution.
Debug Mode — Structured debugging with auto-cleanup. Find and fix issues systematically.
Council Mode — Multi-agent deliberation for complex decisions. Technical review, architecture, and security templates — multiple AI perspectives before you commit.
Text-to-application — Generate full-stack web apps and API services from a description. Turn a research paper, RFC, or spec into working code.
Multi-agent management — Run multiple agents in the background, auto-detecting CPU cores for optimal parallelism.
Staged changes review — Every AI edit held in the Source Control panel. Accept or reject individual changes or batches.

IDE Capabilities

Council Mode

Multi-agent deliberation for complex decisions. Create custom councils with specialized AI agents. Technical review, architecture, and security templates. Multiple AI perspectives before you decide.

Council Mode
(screenshot)

Agent Backends

TurboLight for fast tasks. AutoAgent for multi-agent orchestration with web browsing and memory. DeepCode for research paper-to-code. FastAgent for deep research with GUI automation.

Calliope AI IDE — an autonomous agent in the side panel summarizing a repository's structure and tech stack next to the code editor

Intelligent Permission System

Three profiles — Conservative, Balanced, Permissive. Approval timeouts, audit logging, sub-agent permission inheritance. Sandbox isolation via Git Worktree or Docker.

Intelligent Permission System
(screenshot)

Staged Changes Review

All AI edits held in Source Control panel. Accept/reject individual changes or batches. Diff viewer for inspection. Group changes by file.

Calliope AI IDE agent panel generating TypeScript code that is staged into the editor for human review before it is applied

Research & Web Integration

Deep research with configurable depth. Web search for current information. Cross-session memory persistence. Computer use / GUI automation.

Calliope AI IDE agent panel rendering a microservices architecture diagram synthesized from research across the codebase

100+ Settings & 50+ Commands

Context-aware suggestions including current file, imports, and workspace structure. Inline chat via right-click or keyboard. Fine-tuned control over every aspect.

Why teams choose it.

The AI IDEs developers know are built to run in one place — their vendor's cloud. That's a non-starter the moment your code can't leave the building. The Calliope AI IDE gives you the same modern AI workflow, and runs inside your own perimeter.

Your models, your keys

9+ providers including local Ollama — bring your own keys, no markup, no lock-in to one vendor's models.

Runs where your code lives

Desktop, Docker or JupyterHub; on-prem or air-gapped. Telemetry off by default — nothing phones home.

Built for serious work

Five operating modes, autonomous background agents, Council-mode deliberation, and every edit staged for human review.

Agent backends & the text-to-application stack.

Four specialized agent backends for different kinds of work, plus full-stack code generation from a description, an RFC, or a research paper.

Agent backends

Backend	Purpose
TurboLight	Fast, lightweight baseline for everyday tasks.
AutoAgent	Multi-agent orchestration with web browsing and memory.
DeepCode	Research paper-to-code transformations.
FastAgent	Deep research with GUI automation.

Text-to-application stack

Target	Frameworks & libraries
Text-to-Web	React, Vue, Svelte, Next.js, Nuxt, Astro — with Tailwind, shadcn/ui, Material UI, Chakra, Bootstrap.
Text-to-Backend	Express, Fastify, NestJS, FastAPI, Django, Flask — over PostgreSQL, MySQL, MongoDB, SQLite, Redis, with JWT / OAuth2 / API-key / session auth and auto-generated Docker + OpenAPI docs.
Paper / Spec-to-Code	Parse research papers or feed in technical specs and get working implementations — from RFC to code, from design doc to codebase.

The most capable AI IDE.

A serious tool for serious work.

The right AI mode for every task — and you always stay in control.