Home

Refact Documentation

Documentation for the local-first Refact coding assistant.

Refact is a local-first coding assistant for VS Code and JetBrains IDEs. The IDE extension starts the local refact-lsp engine, connects it to the model providers or local runtimes you configure, and gives you chat, agent workflows, code completion, project context, and integrations from inside your editor.

Refact does not require a hosted Refact account or bundled model service. You bring your own provider keys, OAuth-capable provider account, or local model runtime.

Start In Five Minutes

1. Install Refact for VS Code or JetBrains IDEs.
2. Open the Refact sidebar or tool window in your IDE.
3. Add at least one provider or local runtime.
4. Choose default chat, reasoning, and completion models.
5. Start a chat, ask the agent to work on a task, or use inline code completion.

For a guided first-run path, follow the Quickstart.

Core Pillars

Local Engine

The Refact engine runs on your machine and communicates with the IDE over localhost. It manages chat sessions, tools, code completion, local indexes, trajectories, tasks, and knowledge.

IDE Chat And Agent Workflows

Use Refact chat to ask questions about your codebase, attach context, and switch into autonomous agent modes for multi-step edits, searches, tests, and integration-backed workflows.

Code Completion

Configure a completion-capable provider or local runtime to get fill-in-the-middle suggestions directly in your editor.

Providers And Local Runtimes

Refact supports BYOK providers such as Anthropic, OpenAI, OpenRouter, Groq, DeepSeek, Gemini, xAI, Qwen, Kimi, Zhipu, MiniMax, GitHub Copilot, Claude Code, and custom OpenAI-compatible endpoints. It also supports local runtimes such as Ollama, LM Studio, and vLLM.

Integrations

Enable the tools you need: shell commands, long-running command services, Chrome, GitHub, GitLab, Docker, PostgreSQL, MySQL, PDB, MCP, and other configured integrations.

Local Context, Knowledge, And Tasks

Refact stores project trajectories, knowledge, task boards, and usage summaries locally. Requests leave your machine only for providers and integrations you configure.

Recommended Next Pages

• Quickstart
• Installation Overview
• Configure Providers
• Supported Models
• Privacy
• Worktrees

Key Features

• AI chat and autonomous agent workflows.
• Code completion.
• Local project trajectories, tasks, knowledge, and worktree isolation.
• BYOK and local provider setup.
• Local token/provider-cost usage summaries.

Refact does not require a hosted account or managed Refact inference service.