Which IDEs does Swylink support?

Swylink supports 7 MCP-capable IDEs: Cursor, Claude Code, Antigravity (Gemini), OpenAI Codex, Windsurf, Cline/RooCode, and GitHub Copilot. Any IDE supporting MCP over stdio can also be connected.

What makes the context intelligent?

Every context snapshot stores structured metadata: a summary, key decisions, topic tags, and files changed. Swylink generates vector embeddings and uses semantic search — so your AI can ask 'what was decided about auth?' and find relevant context by meaning, not keywords.

What happens when I switch between AI tools?

Nothing is lost. Context saved by Cursor is instantly searchable by Claude Code, and vice versa. Each AI automatically saves context after completing tasks and searches before starting new ones — creating a continuous, cross-tool memory.

What is the Model Context Protocol (MCP)?

MCP is an open standard by Anthropic that lets AI tools communicate with external services via a structured JSON-RPC protocol over stdio. Swylink uses MCP to provide save_context and search_context tools to any compatible IDE, enabling persistent memory without IDE-specific plugins.

How does Swylink handle proprietary or confidential code?

Swylink only stores context metadata — not your source code. All data is transmitted via encrypted WebSocket connections and stored with Supabase row-level security, ensuring only you can access your context. Enterprise customers can request dedicated infrastructure.

What vector model does Swylink use for embeddings?

Swylink uses 768-dimensional vector embeddings processed by pgvector in PostgreSQL. Each context snapshot is embedded and stored, allowing cosine similarity search that finds relevant past decisions by meaning rather than exact keyword matches.

How many IDEs does Swylink support?

Swylink supports 7 MCP-capable platforms: Cursor, Claude Code (Anthropic), Windsurf (Codeium), GitHub Copilot, OpenAI Codex, Cline/RooCode, and Antigravity (Google Gemini). Any IDE that supports MCP over stdio can also connect.

What is the latency of context search?

Semantic search queries typically return results in under 200ms. The search happens server-side using pgvector's HNSW index, which provides approximate nearest neighbor search optimized for low-latency responses even with thousands of context snapshots.

How much context can Swylink store?

The Pro plan includes 1,000 messages per month, where each message is a context save or search operation. There is no limit on the total amount of stored context — all your historical context snapshots remain searchable indefinitely.

How do I connect a second AI tool?

Run npx swylink connect again and select your second IDE. Swylink automatically detects the IDE and writes the MCP configuration. Once connected, both tools share the same context workspace — search in one, find decisions from the other.

What happens to my context if I cancel my subscription?

Your stored context remains accessible in read-only mode for 30 days after cancellation. During this period, you can export your data. After 30 days, context snapshots are permanently deleted from our servers.

Is there a team or enterprise plan?

Team and enterprise plans are coming soon. Contact contact@swylink.com for early access to shared workspaces, team context visibility, and dedicated infrastructure options.

New — works with 7 AI tools

Stop explainingtwice

Your AI tools forget everything between switches. Swylink gives them shared memory — what one learns, they all remember.

Scroll to explore

Works with 7 AI coding tools

Cursor

Claude Code

Windsurf

GitHub Copilot

Antigravity

OpenAI Codex

Cursor

Claude Code

Windsurf

GitHub Copilot

Antigravity

OpenAI Codex

Cursor

Claude Code

“Refactored auth from JWT to session-based storage”

architectureauthmigration

src/auth/session.ts · src/middleware.ts

Claude Code

Windsurf

“Added rate limiting middleware — 120 req/min per user”

apisecurityperformance

src/middleware/rate-limit.ts

Windsurf

Copilot

“Chose pgvector over Pinecone for embeddings — lower latency”

databasedecisionsearch

src/context/embedding.service.ts

Shared memory in action

What one AI learns,they all remember

Cursor refactors your auth. Thirty seconds later, Claude Code already knows. Decisions, files, reasoning — every AI finds what it needs by meaning, not keywords.

Instant

cross-tool sync

AI tools connected

<1 min

to set up

How it works

Three steps. Sixty seconds.

Give your AI tools shared memory in under a minute. No technical setup — your AI handles everything.

Pick your AI

Paste one message

Swylink generates a setup message. Paste it in your AI chat — it configures the connection automatically.

They remember everything

Your AIs start sharing memory instantly. Connect a second tool and they know what each other decided, built, and learned.

Pricing

One plan. Unlimited AIs.

Everything you need to give your AI tools shared memory.

Pro

From $9.00/mo

Local currency shown at checkout when available.

Unlimited AI workspaces
1,000 messages/month
Priority support

FAQ

Still not sure?

Is my data private?

The Swylink CLI runs locally on your machine and connects via encrypted WebSocket. Context is stored per-user with row-level security — no other user can access your data. The backend never reads or processes your code files directly.

Does Swylink access my source code?

No. Swylink captures context snapshots — summaries of what happened, key decisions made, topic tags, and file paths changed. It does not read, store, or transmit your actual source code. The AI in your IDE decides what context to save based on the conversation.

How much does Swylink cost?

Swylink offers a Pro plan with unlimited AI workspaces and 1,000 context messages per month. Local currency is shown at checkout when available. You can start with a free trial to test the full experience.

How does Swylink differ from Cursor's built-in memory?

Cursor's memory is session-scoped and stays within Cursor. When you switch to Claude Code or Windsurf, that context is lost. Swylink creates a shared memory layer that persists across all 7 supported IDEs via MCP. Decisions made in Cursor are instantly searchable in Claude Code, and vice versa.

What is Swylink?

Swylink is an intelligent context layer for AI coding tools. It gives Claude Code, Cursor, Codex, Windsurf, and other MCP-enabled IDEs persistent, searchable memory — so every AI in your stack knows what the others decided, built, and learned.

How do I connect my IDE?

Create a free account, pick your IDE, and paste the generated prompt into your AI chat. The AI configures everything automatically — no manual file editing needed.

How does semantic search work?

When you save context, Swylink generates a 768-dimensional vector embedding of the summary. When an AI searches, the query is embedded and matched against stored contexts using cosine similarity in pgvector — returning the most semantically relevant results.

Can I use Swylink offline?

The CLI requires an internet connection to sync context with the Swylink cloud and perform semantic search. However, your IDE continues to work normally offline — context syncing resumes automatically when you reconnect.

Your AIs have amnesia.

Every time you switch tools, your AI forgets everything you told it. Give them shared memory — and never explain yourself twice.

5 free messages. No credit card.