# Reflexor Reflexor is the local-first, zero-cloud AI coding-assistant and Copilot alternative for VS Code, Cursor, Antigravity, and VSCodium. Powered by Ollama, DeepSeek, DeepSeek-R1, Qwen, LM-Studio, and Llama3, Reflexor delivers private local-ai and local-llm inference, fast code-completion, embedded codebase-chat, and approval-first agentic workflows right on your machine. [Install for VS Code](https://marketplace.visualstudio.com/items?itemName=reflexor-app.reflexor) | [Install for Cursor](https://open-vsx.org/extension/reflexor-app/reflexor) ![Reflexor brand banner](https://reflexor.com.br/og-image.png) ## Why developers use Reflexor ### Reflexor vs Cloud AI Assistants | Capability | Reflexor | Cloud Copilot / Cloud Cursor | | :-------------------------- | :--------------------------------------------------------------------------- | :------------------------------------------------------------------ | | **Code Privacy & Security** | **100% Local-first & airgap ready** (zero-cloud privacy-first architecture) | Code, prompt context, and file buffers sent to remote cloud servers | | **Cost & Licensing** | **$0 Free to use** (No monthly subscription or hidden fees) | $10 to $20 per seat every month | | **Local RAG & Embeddings** | **In-process embedded Rust engine** (zero Docker, zero external vector DB) | Remote cloud indexing or heavy external background databases | | **Model Freedom** | **Ollama, DeepSeek-R1, Qwen, LM-Studio, Llama3, MCP** | Locked proprietary vendor models | | **Agent Safety & Control** | **Approval-first code-review** (Every file diff and command requires review) | Blind automated edits and unreviewed script execution | | **Developer Environments** | **VS Code, Cursor, Antigravity, Windsurf, VSCodium + standalone native CLI** | Single editor plugin only | ### Core Developer Features - **Codebase-chat & semantic-search**: Chat with your codebase, ask questions about the active file or selection, attach files, and retrieve indexed workspace context using local embeddings. - **Embedded local RAG**: Search and retrieve project context with an in-process Rust sidecar that generates and persists vector embeddings locally on your hardware. - **Inline-completion & autocomplete**: Ghost suggestions while you type code-completion, without transmitting your buffer to any external cloud API. - **Approval-first ai-agent & coding-assistant**: Propose precise refactoring and shell commands, then inspect every diff before an agentic action commits. - **Hardware-aware setup**: Reflexor inspects your CPU and GPU to recommend optimal local models (DeepSeek, Qwen, Llama3) with one-click guided download. - **Native CLI & IDE workflow**: Run Reflexor seamlessly inside VS Code, Cursor, Antigravity, and VSCodium, or pair it with the companion CLI for terminal workflows. - **Model Context Protocol (MCP)**: Discover, connect, and invoke MCP tools locally to expand agent capabilities safely. - **Optional web research**: Retain an airgap offline-ai workflow by default, or grant one-time consent for web search when public information is required. ## Install and first run 1. Open **Reflexor Chat**. 2. If the local engine is missing, choose **Set up local engine** and confirm the one-time download. 3. When Reflexor recommends models, choose **Continue and download**. Model weights are never downloaded silently. 4. Use **Ask** for a question, **Plan** for a written approach, or **Agent** to propose an edit. Review any change before it lands. You do not need to learn model names or configure a terminal. Chat works after the engine and a model are in place. Workspace indexing is optional. ## See it in action These screenshots show the core workflow in a real editor: local codebase-chat, inline-completion, approval-first ai-agent review, and first-run engine setup. ### Codebase chat and local context ![Reflexor codebase chat using local workspace context](https://reflexor.com.br/readme/reflexor-chat-context.png) Ask about the active file, selected code, or indexed workspace evidence with semantic-search without leaving the editor. ### Inline completion and autocomplete ![Reflexor inline code completion and autocomplete in the editor](https://reflexor.com.br/readme/reflexor-inline-completion.png) Get local code-completion, inline-completion, and AI autocomplete while you work. ### Agent proposal, code-review, and refactoring ![Reflexor Agent proposal with reviewable refactoring diff](https://reflexor.com.br/readme/reflexor-agent-review.png) The coding-assistant and ai-agent workflows propose changes first for code-review. Review the diff before an agentic coding action is applied. Agent workflows are experimental. ### Local engine and Ollama setup ![Reflexor local engine and Ollama model setup](https://reflexor.com.br/readme/reflexor-setup.png) The setup flow explains the local engine, Ollama integration, model roles, and downloads before anything large is installed. ## Before you install - Open only trusted local folders. Untrusted and virtual workspaces are not supported. - The VSIX does not ship the engine or model weights. Those downloads happen only after you confirm. - CPU-only machines are supported. Speed depends on the model and your hardware. - Local chat, completion, and indexing do not need a cloud account. - **Agent** can propose file writes and shell commands. Those stay experimental and need your approval. Reflexor is free to use and proprietary. Source code is maintained privately and is not distributed under an open-source license. ## Supported editors and platforms - Visual Studio Code - Cursor through this Open VSX listing - Google Antigravity & Antigravity IDE - Windsurf & compatible Codeium builds - Compatible VSCodium and VSCode builds - Windows x64 and Windows ARM - Linux x64 and Linux ARM64 - macOS Apple Silicon and macOS Intel, each in its own thin package The listing shows the editor version it needs. The current package requires VS Code `^1.106.0` or a compatible Cursor, Antigravity, or Windsurf build. Install the VSIX that matches your operating system and CPU. Linux ARM64, Windows ARM, and macOS Intel each have their own package. ## Models and local engine Reflexor can: - reuse a system Ollama or LM-Studio installation when one is already available - download and manage its embedded local engine after you confirm - recommend chat and embedding models based on available hardware - use local embeddings for richer codebase retrieval The VSIX does not ship model weights or an Ollama binary. Those downloads happen only after confirmation. Model licenses belong to their providers, not to Reflexor. Speed and quality depend on the selected LLM, CPU, memory, optional GPU, and project size. CPU-only machines are a supported mode. ## Chat, context, and Agent Use **Ask** for explanations, **Plan** for a written approach, and **Agent** for engineering workflows that propose edits or commands. Reflexor is a developer tool for local-first workspace AI, not an autonomous writer. Reflexor can use: - the selected code - the active file - files you explicitly add to chat - local workspace indexing, RAG, semantic-search, and code search - diagnostics and project structure when the workflow needs them Review the proposed diff before accepting it. Agent workflows are experimental and can still produce incorrect plans or code. Session auto-approve is off unless you turn it on. Policy, path, and revalidation checks still apply. ## Web Search, Web Research, and remote inference The core product is local-first. Web access follows one policy, shared with the Reflexor CLI. - Web search runs without asking at every approval level. It sends only the sanitized query text; never code, file contents, absolute paths, or RAG context. - Opening a page asks per domain: once, for this conversation, always, or no. "Always" saves the domain in `reflexor.web.alwaysAllow`; `reflexor.web.alwaysDeny` blocks a domain for good. Pages a search opens to ground its answer (up to 3) do not ask. - Localhost, private network, link-local and cloud-metadata addresses are always blocked. - Switch it off with `reflexor.web.mode` (`off`, or `ask` to be asked before searching too), `reflexor.webSearch.enabled`, or `REFLEXOR_NO_WEB=1`. - Every web access is recorded in a local redacted log (`web-access.jsonl` under `~/.reflexor/logs`). It never leaves your machine and never stores page content. - Pointing inference at a non-local Ollama endpoint requires a separate consent. Prompts and embeddings then go to that endpoint, not to Reflexor. ## Privacy See the full [privacy policy](https://reflexor.com.br/en/privacy/). - Core chat, inline-completion, project indexing, and local embeddings run on your computer. - Conversation history and prompts stay local for local-engine turns. - Privacy-first private AI and private coding workflows are the default when you use the local engine. - Offline AI and offline coding work when the local engine and required models are installed. - An airgap session uses the installed local engine with remote inference left off. - Product telemetry is off by default and, when enabled, remains on your machine. - You can export or delete local conversations from the Command Palette. Reflexor does not claim to be fully offline or private under every configuration. Read the privacy policy before enabling web features, remote inference, or integrations. ## Requirements - Visual Studio Code, Cursor, or Antigravity that meets the version shown on the listing - Windows x64, Linux x64, or macOS Apple Silicon - A chat or coder model, installed only after confirmation - Optional embedding model such as `nomic-embed-text` for richer workspace retrieval - Linux: `tar` with zstd support, or a `zstd` binary, for the embedded engine archive - macOS: Apple Silicon (`darwin-arm64`). Metal-capable hardware is recommended when available ## Troubleshooting - **Backend offline**: Reload the window and check **Output: Reflexor**. - **Local engine offline**: Wait a few seconds, then reopen Reflexor Chat and use **Set up local engine**. - **No models**: Use **Continue and download** in the Reflexor setup checklist. - **Slow completion**: Choose a smaller model and avoid oversized context overrides. - **Stale project context**: Rebuild the local project index after large refactors. - **Need a bug report**: Run **Reflexor: System Diagnostics**, redact secrets, and email the excerpt. Agent workflows are experimental. Performance varies with model, hardware, context size, and project size. ## Support and legal - Email: [contato@reflexor.com.br](mailto:contato@reflexor.com.br) - Website: [https://reflexor.com.br](https://reflexor.com.br) - [Support](https://reflexor.com.br) - [Privacy policy](https://reflexor.com.br/en/privacy/) - [Terms of use](https://reflexor.com.br/en/terms/) - [Security policy](mailto:contato@reflexor.com.br) Feedback after a completed workflow is welcome, whether the result was good or not. Reflexor never requires a rating or a positive review. ## License Reflexor is **proprietary** software. The package license and legal terms are described on the [Reflexor terms page](https://reflexor.com.br/en/terms/). The complete license text is included in the installed package. Third-party components remain under their own licenses and notices included in the installed package. Source code is maintained privately and is **not** open source.