Give Cline intelligent cross-tool context
Swylink gives Cline and RooCode persistent memory from your entire AI stack. Structured decisions, tags, and semantic search keep everything aligned.
Connect Cline / RooCode in 4 steps
- 01Installnpx swylink@latest init
- 02Authenticatenpx swylink auth
- 03Connectnpx swylink connect
- 04CodeOpen Cline / RooCode & go
Where Swylink writes the Cline / RooCode config
~/.cline/data/settings/cline_mcp_settings.jsonRunning npx swylink connect automatically detects Cline / RooCode and writes the MCP bridge configuration to this path. No manual editing required.
Why Cline and RooCode need persistent cross-tool memory
Cline and RooCode are VS Code extensions that bring a unique plan-and-execute approach to AI-assisted development. Before writing any code, Cline creates a detailed plan of what it intends to do, then executes each step with your approval. This deliberate, transparent workflow generates incredibly rich reasoning artifacts — the plan itself, the rationale for each step, the alternatives considered, and your approval decisions along the way.
The problem is that this valuable reasoning context is confined to the current Cline session within VS Code. When you close the session or switch to Claude Code for a task that requires deeper reasoning, all of Cline's planning artifacts disappear. The detailed step-by-step rationale that Cline generated — which could inform future decisions across any AI tool — is lost.
Swylink captures the reasoning behind Cline's multi-step plans and persists it as searchable context. Each approval decision, each step's rationale, each alternative that was considered and rejected — all stored with structured metadata and semantic embeddings. When you open Cursor or Claude Code later, you can search "how did we plan the database migration" and retrieve the full reasoning chain from Cline's execution. RooCode users benefit equally, as both extensions use the same MCP integration path.
Context that flows from Cline to your other AI tools
Cline's plan-and-execute model generates uniquely structured context that Swylink captures: step-by-step execution plans with the reasoning behind each step, approval decisions where you guided the AI's direction, alternatives considered and rejected during planning, incremental implementation rationale as each step built on the previous, file-by-file change reasoning within multi-step plans, and rollback decisions when steps were revised mid-execution. This planning context is invaluable for other tools — Claude Code and Cursor gain access to the full deliberation trail.
Frequently asked questions about Swylink and Cline / RooCode
Does Swylink capture Cline's step-by-step plans?
Yes. Swylink captures context from all Cline interactions through MCP, including the planning phase and each execution step. The structured metadata preserves the reasoning chain, so other AI tools can understand not just what was done, but the deliberate plan that led to each change.
Where is the MCP config for Cline?
Cline reads MCP configuration from ~/.cline/data/settings/cline_mcp_settings.json. Running npx swylink connect detects Cline and writes the server block to this path. RooCode uses the same configuration location, so both extensions are connected simultaneously.
Does Swylink work with both Cline and RooCode?
Yes. Cline and RooCode share the same MCP integration path and configuration location. When you connect Swylink to one, it works with both. Context saved during a Cline session is equally accessible from RooCode, and both contribute to the same shared memory that your other AI tools can search.