"How to work with local files" matching MCP servers:

Local-only MCP server for saving and restoring conversation context across AI agent sessions using JSON files on disk.

Privacy-first local document search using semantic search. Runs entirely on your machine with no cloud services, supporting PDF, DOCX, TXT, and Markdown files.

Gives Claude persistent, local context about how you work, your weekly agenda, and shipped work, with automatic work logging and git integration—all stored privately on your machine.

Manages persistent work memories and context across AI tools, enabling structured storage, search, and retrieval of work content, tasks, and knowledge with session-based project organization.

Read-only MCP server for a local-first supply chain knowledge base, enabling knowledge search with source evidence, internal system troubleshooting, vendor documentation fallback, vessel tracking, and work handoff preparation.

Enables saving project facts as they occur and retrieving only task-relevant entries when starting new work, using Jev evaluation to rank them while preserving all stored context.

coldstart gives coding agents durable memory of a specific codebase, with no embeddings, no vector store, and no API key. Agents write notes after real tasks — what a file is for, how a flow spans files, which invariants hold — and those notes are surfaced automatically to a later session whose task matches, so nothing gets rediscovered from scratch. Every note is anchored to real files by content

Your coding agent remembers how it solved past tasks and reuses it next time: trigger, steps, outcome, lessons. Hooks recall the right skill before a task; skills strengthen with confirmed use and unused ones decay. Local SQLite, no API key, nothing leaves your machine. 9 MCP tools, works with Claude Code, Codex and Cursor. Agent-written memory stays untrusted until you approve it.

Self-improving, verifiable memory for AI coding agents. Learns how you work, stops repeating mistakes, models each project, recalls the right lesson at the right moment. Every memory is signed and tamper-evident. Local-first.

Semantic memory for AI builders: capture the tacit engineering know-how that never reaches your docs, recall it the moment it applies. Built in Rust on Postgres and pgvector.

Simple, fast, long-term memory for your AI chatbots, agents, and MCP tools. Give your LLM a persistent memory using a local SQLite file. No complex database servers, no heavy cloud setup, and zero C++ compilation errors. Works everywhere.

Local MCP server that manages a job search using Markdown files, enabling skills tracking, gap analysis, and offer pipeline management via Claude.

Provides an AI agent with a searchable markdown notes vault, offering tools to create, read, list, search, update, and delete notes stored as plain .md files.

Model-agnostic, agent-ready Obsidian MCP server with RBAC, SLSA provenance, and native search. 163 tools across 31 domains, multi-vault, pluggable embeddings.

A persistent memory MCP server for AI assistants, storing and retrieving notes using local JSONL files with BM25 search.

Local MCP runtime that gives persistent cross-session memory to 12 AI coding agents (Claude Code, Cursor, Codex, Gemini CLI, Windsurf, Amp, and more). SQLite-backed state survives context resets. Install: npm install -g @egchq/egc

Persistent local semantic memory for MCP tools, enabling storing and searching notes by meaning using local embeddings and SQLite.

Enables coding agents to search a local, offline archive of past sessions and repository Markdown docs using hybrid lexical and semantic retrieval, and read matching records to ground new work in prior knowledge without external calls.

A local MCP server that saves work-in-progress checkpoints to SQLite, enabling AI agents to resume interrupted sessions without redoing work.

Scans local folders for PDF papers, extracts title/abstract text, and moves files to categorized folders based on classification decisions made by Claude or another MCP server.

A codebase-memory MCP server for AI coding agents with hybrid semantic+BM25 code search, structural outlines, a work journal, and gaze-and-think deep reasoning over a local LanceDB database.

A local-first MCP server that exposes personal notes and files as unified semantic context for AI agents via vector search and file monitoring.

A local-first MCP server that gives AI coding assistants persistent, structured, human-readable memory for a software project by storing project knowledge as Markdown files in the project's .dev-context-memory/ folder.

Knowledge-lifecycle MCP: turn agent work into memory, recall it across projects

Search past OpenCode conversation history before starting new work on a module or file, via a local read-only FTS5 index built from OpenCode's own SQLite database. No network calls, fully local. 7 tools for keyword search, file lookup, and session browsing.

Search firsthand observations that agents recorded while doing real work — what actually happened with a product, API, service, or place, rather than what its documentation claims. Agents can also write back what they observed.

Your AI forgets everything between sessions. Rekindle fixes that. For Claude Code users who lose time re-explaining project context every session.

StelaSpace is an HTML-first knowledge base for the artifacts AI agents create. This MCP server lets Claude Code, Claude Desktop, and Codex publish HTML files straight from your editor — it takes a file path and uploads from disk, so even large data-heavy reports work. Free tier available.

Project Memory stores decisions, evidence, outcomes and reviewed lessons in local SQLite. Assistants retrieve bounded context through three MCP tools. People follow and plan work in a live workspace with an overview, sprint board, decision history, readable documents, skills and project maps. Offline exports remain available.

Standard coding agents waste 70% of context tokens re-reading raw files with shell grep. Synapse-MCP indexes your codebase into a sub-millisecond local AST graph—giving Claude Code, Cursor, and Windsurf exact caller trees, 60% lower token spend, and zero-hallucination refactoring.