# Swylink > Swylink is an intelligent context layer for AI coding tools. It gives Cursor, Claude Code, Codex, Windsurf, Antigravity, Cline, and Copilot persistent, searchable memory — so every AI in your stack knows what the others decided, built, and learned. **Version:** 1.0 | **Updated:** 2026-04-07 | **License:** RSL 1.0 | **Pricing:** Free trial, Pro from $9/mo Swylink solves the "context amnesia" problem: when developers switch between AI coding assistants, each tool starts from zero. Swylink gives them intelligent context via MCP — with semantic search (768-dim vector embeddings), structured metadata (decisions, tags, files), and automatic context saving/searching through IDE rules injection. Swylink differs from built-in IDE memories because it synchronizes context across multiple AI tools simultaneously via MCP. When a developer makes architectural decisions in Claude Code, those decisions are instantly searchable in Cursor, Windsurf, or any other connected IDE. ## How it works 1. Install the CLI: `npx @swylink/cli` 2. Authenticate: `npx @swylink/cli auth login --api-base-url https://api.swylink.com` 3. Generate a one-time setup token in Dashboard -> Connections 4. Start the bridge: `SWYLINK_SETUP_TOKEN= npx -y @swylink/cli start --server wss://api.swylink.com/ws --ide ` 5. Diagnose issues: `npx @swylink/cli doctor --ide ` ## Supported IDEs (7 platforms) - Cursor — config: `.cursor/mcp.json` - Claude Code (Anthropic) — config: `.mcp.json` - Antigravity (Gemini) — config: `~/.gemini/antigravity/mcp_config.json` - OpenAI Codex — config: `~/.codex/config.toml` - Windsurf — config: `~/.codeium/windsurf/mcp_config.json` - Cline / RooCode — config: `~/.cline/data/settings/cline_mcp_settings.json` - GitHub Copilot — config: `.vscode/mcp.json` - Any IDE supporting MCP over stdio ## Glossary - **Context snapshot**: A structured record captured after each AI coding session containing a summary, key decisions, topic tags, and files changed. - **MCP bridge**: The WebSocket connection between the Swylink CLI and the cloud that enables real-time context sync across IDEs. - **Setup token**: A one-time bootstrap credential used during initial IDE connection. After first use, Swylink stores a local device credential. - **Semantic search**: Finding past context by meaning rather than keywords, using 768-dimensional vector embeddings with cosine similarity via pgvector. - **Rules injection**: Automatic configuration of IDE-specific rules files so each AI proactively saves and searches context without manual work. ## FAQ - **What is Swylink?** Swylink is a persistent context layer that uses the Model Context Protocol (MCP) to synchronize memory across 7 AI coding tools. It captures structured metadata (decisions, tags, files) and enables semantic search across all connected IDEs. - **How much does Swylink cost?** Free trial available. Pro plan starts at $9/month with unlimited AI workspaces and 1,000 messages/month. - **Is my data private?** Yes. The CLI runs locally, communicates via encrypted WebSocket, and all data is protected with Supabase row-level security. Setup tokens are one-time credentials. - **How does semantic search work?** Each context snapshot is converted to a 768-dimensional vector embedding. When you search, Swylink uses cosine similarity in pgvector to find relevant past decisions by meaning, not keywords. ## Docs - [CLI Getting Started](https://swylink.com/en/docs/cli/getting-started): Install the CLI, authenticate in your terminal, generate a setup token in the dashboard, and connect your IDE with guided bootstrap. - [Install and Verify](https://swylink.com/en/docs/cli/install): The CLI can run through npx, which keeps onboarding light for MVP users and avoids stale global installs. - [Login from the Terminal](https://swylink.com/en/docs/cli/login): Login uses your Swylink account credentials in the terminal, stores a local session file, and keeps the password out of subsequent CLI commands. - [Generate Setup Tokens](https://swylink.com/en/docs/cli/api-keys): Create one-time setup tokens in Dashboard -> Connections. The CLI exchanges them for local device credentials and reuses them automatically. - [Connect Your IDE](https://swylink.com/en/docs/cli/connect-ide): Each IDE uses a slightly different MCP config file, but the bridge command stays consistent: run Swylink over stdio with a setup token and an explicit IDE identifier. - [Command Reference](https://swylink.com/en/docs/cli/commands): These are the operational commands you need for day-to-day CLI use. - [Troubleshooting](https://swylink.com/en/docs/cli/troubleshooting): When onboarding fails, the issue is usually session state, the MCP config path, or a wrong server URL. Check these in order. ## Integrations - [Cursor Integration](https://swylink.com/en/integrations/cursor): Setup guide and FAQ for Swylink with Cursor - [Claude Code Integration](https://swylink.com/en/integrations/claude-code): Setup guide and FAQ for Swylink with Claude Code - [Windsurf Integration](https://swylink.com/en/integrations/windsurf): Setup guide and FAQ for Swylink with Windsurf - [GitHub Copilot Integration](https://swylink.com/en/integrations/github-copilot): Setup guide and FAQ for Swylink with GitHub Copilot - [OpenAI Codex Integration](https://swylink.com/en/integrations/openai-codex): Setup guide and FAQ for Swylink with OpenAI Codex - [Cline / RooCode Integration](https://swylink.com/en/integrations/cline): Setup guide and FAQ for Swylink with Cline and RooCode - [Gemini Antigravity Integration](https://swylink.com/en/integrations/gemini): Setup guide and FAQ for Swylink with Gemini Antigravity ## Blog - [Why AI Tools Forget Context](https://swylink.com/en/blog/why-ai-tools-forget-context): The context amnesia problem and how persistent memory solves it - [What Is MCP](https://swylink.com/en/blog/what-is-model-context-protocol): Developer guide to the Model Context Protocol - [Semantic Search AI Memory](https://swylink.com/en/blog/semantic-search-ai-memory): How vector embeddings power AI memory and context retrieval - [Cross-IDE Context Engineering](https://swylink.com/en/blog/cross-ide-context-engineering): Making your AI tools work together with shared context - [Swylink vs Built-in IDE Memory](https://swylink.com/en/blog/swylink-vs-built-in-ide-memory): Why cross-tool context matters more than single-IDE memory ## Optional - [Full documentation for LLMs](https://swylink.com/llms-full.txt): Complete inline documentation — all CLI docs rendered as markdown in a single file - [About Swylink](https://swylink.com/en/about): Company mission, founder, and product vision - [Website](https://swylink.com): Landing page with product overview and FAQ - [Sitemap](https://swylink.com/sitemap.xml): All indexed pages - [Contact](mailto:contact@swylink.com): Support email