# Supercode

> An open-source AI-powered SWE agent and CLI coding assistant that ships code with you, supporting Claude, GPT, Gemini, and 15+ model providers.

Supercode is an open-source AI software engineering (SWE) agent built as a CLI tool and full-stack platform, licensed under MIT and sponsored by Vercel. It lets developers write, review, and ship code using natural language commands, with support for free included models or any external provider. The project is currently in beta and live on Product Hunt, with 212 GitHub stars and 56 releases shipped as of the latest changelog.

## What It Is

Supercode is a terminal-first AI coding agent that functions as a pair programmer directly in your shell. It is published to npm as `supercode` and installable via curl, npm, bun, npx, or brew. The core agent supports multi-provider model routing — developers can connect Claude, GPT, Gemini, DeepSeek, Groq, Mistral, Ollama, LM Studio, GitHub Models, NEAR AI, Xiaomi MiMo, LiteLLM, OpenRouter, and Concentrate AI, or switch models mid-session. The platform is structured as a Bun + Turborepo monorepo containing a web dashboard, MDX documentation site, terminal web client, CLI agent, and a parallel open-model fine-tuning project.

## Architecture and Stack

Supercode is built as a full-stack monorepo with clearly separated concerns:

- **CLI agent** (`supercode` on npm) — Bun-powered, uses AI SDK v6 for unified provider access, Commander for argument parsing, and Clack/Chalk for terminal UI
- **Web dashboard** — Next.js 16 with React 19, GitHub OAuth via Better Auth, Inngest for background jobs, and Prisma 7 over PostgreSQL
- **Terminal web client** — A browser UI that mirrors the CLI experience, intended for deployment at `terminal.supercli.com`
- **Cortex SDK** — A developer-facing SDK for building custom AI tooling on top of Supercode
- **Open Model project** — A parallel fine-tuning effort targeting Qwen3-8B and GLM-4-9B via LoRA/QLoRA for coding-specialized LLMs

The agent ships with a tool system covering file read/write, command execution, and search, plus a skills management system for adding, removing, and listing agent capabilities.

## Key Products

The Supercode platform ships three distinct products alongside the core CLI:

- **Supercode Agent** — The primary AI pair programmer that autonomously ships code
- **Supercode Review** — Automated code review triggered on every pull request
- **Supercode Voice Agent** — Hands-free, voice-controlled development for coding without a keyboard

## Integrations and Partnerships

Supercode has published partnerships with three infrastructure providers:

- **Concentrate AI** — LLM API gateway providing the model routing backbone, with a claimed 99.9% LLM API uptime
- **Merge** — Connects Supercode's AI coding agents to 200+ developer tools through Merge's Agent Handler and Unified API
- **Dodo Payments** — Handles subscription billing, checkout, recurring payments, and license fulfillment

The platform also integrates with Pinecone for vector storage, Prisma for database ORM, and supports GitHub OAuth for authentication.

## Update: v0.1.90 and Recent Releases

The homepage changelog shows v0.1.90 as the current version, with 56 total releases shipped. The latest GitHub release tag is v0.1.65, published in July 2026, which introduced MCP integration, chat enhancements, and Composio support. Recent notable changes include:

- Claude Opus 5 added as a supported model
- New skills management commands (add, remove, list agent skills)
- Hallucination guard in the agent runner — detects when a model claims actions without actual tool calls
- Usage event migration with a new `usage_event` table for detailed tracking

The repository was last pushed in September 2026, indicating active development. The open-model fine-tuning track (Qwen3-8B and GLM-4-9B) is listed as in-progress.

## Features
- AI pair programmer CLI agent
- Automated code review on pull requests
- Voice-controlled coding (hands-free development)
- Multi-provider model support (Claude, GPT, Gemini, DeepSeek, Groq, Mistral, Ollama, and more)
- Switch models mid-session
- Free models included
- Skills management system (add, remove, list agent skills)
- Hallucination guard in agent runner
- MCP integration
- Cortex SDK for custom AI tooling
- Web dashboard for repository management and analytics
- Terminal web client (browser-based CLI mirror)
- GitHub OAuth authentication
- File read/write and command execution tools
- Open-source LLM fine-tuning (Qwen3, GLM-4)

## Integrations
Claude (Anthropic), OpenAI GPT, Google Gemini, DeepSeek, Groq, Mistral, Ollama, LM Studio, GitHub Models, NEAR AI, Xiaomi MiMo, LiteLLM, OpenRouter, Concentrate AI, Merge Unified API, Dodo Payments, Pinecone, Composio, Vercel, GitHub OAuth, Inngest, PostgreSQL, Prisma

## Platforms
CLI, WEB, API, DEVELOPER_SDK

## Pricing
Open Source

## Version
v0.1.90

## Links
- Website: https://www.supercodeai.tech
- Documentation: https://supercli-docs.vercel.app/
- Repository: https://github.com/yashdev9274/superCli
- EveryDev.ai: https://www.everydev.ai/tools/supercode