Getting Started

LocalNest MCP is a local-first MCP server that gives AI agents scoped access to your codebase, with optional semantic indexing, hybrid retrieval, and local memory.

Documentation

Reference-first docs for installing, indexing, searching, and operating LocalNest.

This site is structured for implementation work. Start with setup, then move into search and indexing, and use release notes only when you need version-specific behavior.

Current stable package0.0.5

Preferred backendsqlite-vec

Minimum runtimeNode.js 18+

Quick start

If you want the shortest path:

1. Install globally.
2. Run setup.
3. Run doctor.
4. Paste the generated MCP block into your client.

Recommended global install:

npm install -g localnest-mcp
localnest install skills
localnest setup
localnest doctor

Fallback:

npx -y localnest-mcp-setup
npx -y localnest-mcp-doctor

Upgrade paths:

localnest upgrade
localnest upgrade stable
localnest upgrade beta
localnest upgrade 0.0.5

Choose your path

• New install: start with Install, then Configuration.
• Architecture and internal flow: use Architecture.
• Daily coding workflow: jump to Tools and Search.
• Version-specific behavior: use Current release and Release matrix.

New machineInstall the package, skill, and generated MCP config with the current stable defaults.Current releaseSee what changed in 0.0.5 before you upgrade or publish.Archived beta.8Open the full archived doc set when you need the previous-beta behavior.

What you get

Scoped workspace access

Browse roots, inspect project trees, and read bounded file ranges without exposing unrelated files.

Fast lexical search

Use localnest_search_files and localnest_search_code for modules, exact symbols, and regex patterns.

Local semantic retrieval

Index a project locally, then use hybrid retrieval to answer concept-level questions with better recall.

Durable local memory

Enable opt-in memory capture to persist decisions, preferences, and prior fixes on your machine.

Suggested reading order

InstallSet up the package, skill, doctor checks, sync option, and MCP client block.ConfigurationReview root resolution, environment variables, and backend settings.ArchitectureSee boot flow, retrieval diagrams, indexing, and memory graph behavior.ToolsSee the full tool surface and the intended retrieval workflow.Current releaseSee the 0.0.5 stable behavior, including runtime updates, install behavior, and the current memory workflow.Release matrixCompare stable and beta docs when you are working across versions.

Typical workflow

1

Bootstrap the environment

Install the package, run localnest setup, opt into memory if needed, then verify with localnest doctor.

2

Wire the MCP client

Add the generated mcpServers.localnest block and restart the client.

3

Discover scope

Use localnest_search_files first when locating a feature, module, or directory.

4

Recall prior context

If memory is enabled, run localnest_task_context before deeper analysis so runtime status and relevant memory come back together.

5

Search with intent

Use localnest_search_code for exact identifiers and localnest_search_hybrid for concept retrieval.

6

Validate with exact lines

Read the relevant file window with localnest_read_file before making changes or conclusions.