Paparats – Local Semantic Code Search MCP
Attach local semantic code search across every repo on your machine so the assistant pulls accurate, private context instead of guessing file layouts.
Overview
io.github.IBazylchuk/paparats-mcp is a build-phase MCP server that provides local-first semantic code search across your repos as private context for AI coding assistants.
What is this MCP server?
- Local-first semantic search across all your repositories without sending full trees to the cloud
- MCP server Paparats – Local Semantic Code Search MCP v1.3.1
- streamable-http transport at http://localhost:9876 via @paparats/cli
- Keeps retrieval on your machine for indie builders with many side projects
- GitHub monorepo IBazylchuk/paparats-mcp with server package under packages/server
- Server version 1.3.1
- Default transport URL http://localhost:9876
- npm identifier @paparats/cli
What problem does it solve?
Agents forget where code lives across many local repos and either waste tokens reading whole trees or invent files that do not exist.
Who is it for?
Solo builders with multiple private repos who want grounded agent answers and can run a local HTTP MCP server alongside their editor.
Skip if: Teams that need org-wide hosted code intelligence, strict cloud DLP-only policies without local daemons, or search over non-code knowledge bases only.
What do I get? / Deliverables
After Paparats is running on localhost:9876, your assistant retrieves semantically relevant code from your indexed repos through MCP without exposing the full codebase to a remote indexer.
- Running Paparats MCP server indexed for your local repositories
- Agent queries returning semantically matched code spans as MCP tool results
- Private on-machine context pipeline without mandatory cloud upload
Recommended MCP Servers
Journey fit
Code-search MCP is shelved under build agent-tooling because that is when solo builders connect assistants to the codebase, while the same index also supports review and iteration later. Paparats is marketed as private context for AI coding assistants via a dedicated MCP server, which matches agent-tooling rather than frontend UI or launch distribution work.
How it compares
Local semantic repo search MCP, not a Git hosting integration skill or a generic web RAG crawler.
Common Questions / FAQ
Who is io.github.IBazylchuk/paparats-mcp for?
It is for indie developers and small teams who use AI coding assistants and want fast, private semantic search over all repositories on their own machine.
When should I use io.github.IBazylchuk/paparats-mcp?
Use it whenever your agent needs accurate cross-repo context during implementation, review, or refactoring, especially when code must stay local.
How do I add io.github.IBazylchuk/paparats-mcp to my agent?
Install @paparats/cli from npm, start the Paparats server so it listens on http://localhost:9876, then register io.github.IBazylchuk/paparats-mcp as a streamable-http MCP server in Claude Code, Cursor, or your compatible client.