Companion tools that enhance the daily agentic AI workflow with SPECLAN — the VS Code extension that turns structured specifications into AI-ready prompts.
SPECLAN sits between business intent and AI coding agents as the missing middle layer. It manages specification trees (goals, features, requirements, scenarios, tests) as plain Markdown files in your project folder — no database, no server, no lock-in. These tools extend that workflow beyond the extension itself.
A stdio-to-HTTP bridge that exposes SPECLAN's 42+ MCP tools to any AI assistant that speaks the Model Context Protocol over stdio — such as Claude Code, Cursor, or Windsurf.
The SPECLAN VS Code extension ships with an HTTP-based MCP server (http-mcp). Many AI coding agents, however, expect MCP servers to communicate over stdio. This bridge translates between the two: it connects to SPECLAN's HTTP MCP endpoint and re-exposes every tool via stdio transport.
AI Agent <──stdio──> speclan-mcp-bridge <──HTTP──> SPECLAN VS Code Extension
npm install
npm run build:mcp-bridgeThis produces a self-contained bundle at dist/speclan-mcp-bridge.mjs.
Add the bridge to your AI agent's MCP config. For Claude Code, add to ~/.claude.json or your project's .mcp.json:
{
"mcpServers": {
"speclan": {
"command": "node",
"args": ["/absolute/path/to/dist/speclan-mcp-bridge.mjs"]
}
}
}The bridge connects to http://localhost:8085 by default. Override with the SPECLAN_MCP_URL environment variable:
{
"mcpServers": {
"speclan": {
"command": "node",
"args": ["/absolute/path/to/dist/speclan-mcp-bridge.mjs"],
"env": {
"SPECLAN_MCP_URL": "http://localhost:9000"
}
}
}
}Run directly from source without bundling:
npm run mcp-bridge- Node.js >= 20
- The SPECLAN VS Code extension running with its HTTP MCP server active
MIT
