MCP servers that ship as code — a repository or a package you install and run locally (usually over stdio). They have no hosted endpoint, so we cannot probe or grade them: instead each one shows its repository, install command, license, stars and when it was last updated.
Collected from GitHub repositories and npm, PyPI, Cargo, Go, Maven, NuGet and RubyGems packages tagged as MCP servers, the official MCP Registry, Docker's MCP catalog, ModelScope, Baidu MCP World, Cline and other public directories; one entry per repository. A server that also offers a hosted endpoint is listed under Remote instead. These entries are not counted in our reachability, verification or readiness statistics.
-
Markdown memory for AI agents. Files you can read, edit, grep, and commit. Not a database. Listed in the MCP Registry.
npm MIT ★ 1 -
Causal project memory for coding agents: corrections first in every session, and a guard that blocks the edit that writes a withdrawn value back. 4 MCP tools, ~670 tokens. Claude Code · Codex · Cursor · GitHub Actions
MIT -
Self-hosted knowledge-graph memory for AI assistants. Editable entities and relationships over MCP, with hybrid dense and sparse retrieval.
Apache-2.0 -
context-keepergithub.com/jarmstrong158/context-keeperupdated this year
MCP server that maintains project memory across Claude conversations — records decisions, pipelines, and constraints in human-editable JSON. Zero dependencies.
MIT ★ 1 -
Give any agent self-improving memory you can trust in production. Every correction is kept, shown to your agent when it matters and enforced on Claude Code and Codex. Built-in memory security: notes cannot instruct the agent, and any edit to the record sh
npm MIT -
Long-term memory for AI agents: hybrid-searched memories over human-editable markdown notes. 100% local — offline ONNX embeddings, SQLite FTS5, MCP server, REST API, CLI, dashboard. No cloud, no API keys.
npm ISC -
Local-first AI memory you can see, edit, and override — portable across Claude Code, Codex, Cursor, Windsurf, and other MCP coding tools.
pypi AGPL-3.0 ★ 166 -
Norngithub.com/samad001z/Nornupdated this year
Persistent, local-first memory for AI coding agents. Give Claude Code and Cursor memory across sessions over MCP — and see and edit exactly what your AI remembers. Open source, no API key, no cloud.
MIT ★ 3 -
Lossless local memory for Claude Code. The full, unabbreviated version. Every tool call, every file edit, every thinking block — stored verbatim on your machine. Persistent memory without tokens.
MIT ★ 16 -
Institutional memory for your codebase. Cogito captures architectural decisions and surfaces them automatically when relevant files are edited.
npm MIT -
Local, persistent memory for AI agents (MCP server). Plain JSON on your disk; fact-level search with an optional local semantic layer; corrections replace old facts; stale-value warnings; secrets kept out; a browser page to read, edit and merge it. One command installs it in Claude Code, Claude Desktop, Codex and Cursor.
npm MIT ★ 8 -
Engineering memory for coding agents: the decisions, fixes and constraints behind your code, with who recorded them, the files and the commit, recalled by every agent on the team in Claude Code, Codex, Cursor, Copilot and 60+ MCP clients. Self-hosted community edition; Recallium Cloud at recallium.ai.
NOASSERTION ★ 58 -
Typed, shared memory for coding agents (Claude Code, Codex, any MCP client), stored as an Obsidian vault you can read and edit. Install: uvx agent-mem-vault
MIT ★ 6 -
Shared long-term memory, document search and a code index for AI agents, linked to your code. Memories point at the files and symbols they're about and surface when an agent edits them. MCP server, Claude Code and Codex plugins, SDKs. Self-hosted on ClickHouse.
pypi Apache-2.0 ★ 71 -
Shared FalkorDB context graph for coding agents: a ~200-token map before the first tool call, memory across sessions, edit claims between agents. 144 Claude Code runs: −25% tokens, −29% tool calls. No LLM calls, no Docker.
MIT -
An MCP server fronting a plain-markdown vault for durable, multi-tenant agent memory. One server process serves many agents concurrently; each tool call carries its own scope (agent, user, …) so files are namespaced on disk while a human can still browse and hand-edit the vault directly with any editor.
-
winzegithub.com/coconutpalm/winzeupdated this year
An agentic AI memory system (mcp server) with semantic search and GUI search/edit tool for Markdown knowledge bases
BSD-3-Clause ★ 5 -
Generate responsive, professional email templates using AI with brand memory, expert design, live preview, and easy editing in MCP-compatible clients.
MIT ★ 3 -
Regression memory for coding agents — seal how your repo behaves, then check over MCP (or in CI) that an edit didn't regress it: pass/drift/regressed/missing. With seal history, bisection, and staleness queries.
npm MIT ★ 5 -
localkeepgithub.com/zuneracodes/localkeepupdated this year
You decide what Claude remembers about you. LocalKeep is a personal timeline you curate—titled, dated, editable banks on your machine—then Claude looks them up via MCP instead of inventing uneditable cloud memory.
MIT ★ 1 -
sverklogithub.com/sverklo/sverkloupdated this year
Local-first repo-memory MCP for coding agents. Best for relationship-heavy edits: symbols, callers, dependencies, diff review, and git-pinned decisions before code changes. Start with a no-write proof receipt; use grep/ripgrep for exact strings. MIT; bund
npm MIT -
agentedgithub.com/frane/agentedno update data
A text editor for LLMs, not humans. Take ed, the line editor that nobody has voluntarily used since about 1975, and rebuild it for an environment where the typing user is a language model. Short verbs, line addresses, no modes, no TUI. What an editor optimises for changes when the user is the model: round trips per task, tokens per command, an editing buffer with a long memory, and an undo tree that remembers the branches the agent abandoned, because that's often where the interesting work was. ## Features - **Fewer round trips.** Read-before-Edit is unnecessary; on conflict the response carries the new content so you reconcile in one call instead of pre-reading every time. - **Branching undo.** Walked-back work stays addressable instead of being thrown away when you pick a different path. - **Three-way merge.** Concurrent agents get a structured conflict response instead of a silent overwrite. - **Atomic batches.** Multi-file refactors run all-or-nothing instead of leaving half-applied state on failure. - **Cross-file moves and regex replace as primitives.** Operations the built-in tools can't express cleanly become single calls. - **Drift detection.** External edits to an open file are folded into the tree instead of being clobbered by the next write. - **Inline diagnostics.** Type errors and lint findings surface on save, not at the next build many edits later. - **Cross-session memory.** Per-file notes persist between sessions and surface inline on the next open. - **Audit log.** Every operation recorded with actor and timestamp, so two agents in one workspace can't argue about who moved the head. ## Multi-workspace (v0.4.0) One MCP server registration handles every project you mention. Tool calls route to the workspace owning the absolute path you pass; each project gets its own SQLite, its own undo tree, its own LSP daemon — lazily spawned on first touch. Desktop hosts (Claude Desktop, Codex Desktop, Cursor) install the server once globally and use it across projects without per-project config. ## Install `ae` must be installed locally; this MCP bundle registers `ae serve` as a stdio server. ```sh brew tap frane/tap && brew install agented # or curl -sSL https://raw.githubusercontent.com/frane/agented/master/install.sh | sh ``` Pure Go, single static binary, Apache 2.0. Source: https://github.com/frane/agented
smithery -
Selvedgeno update data
**Persistent decision memory for AI coding agents.** Selvedge records why code changed, which approaches were rejected, and when decisions deserve another look, so future sessions can retrieve that context before editing. Query a function, database column, API route or dependency instead of reconstructing its history from an old conversation. Claude Code, Codex, Cursor, Copilot, Gemini CLI and Windsurf can connect through MCP. The CLI works independently. Decisions live in a local SQLite store, with no required hosted account and no LLM in the core storage or retrieval path. A saved explanation is testimony supplied by a person or agent. Selvedge does not extract hidden model reasoning, verify that an explanation is true, or guarantee that an agent will consult it. [Compatibility and capabilities](https://selvedge.sh/reference/compatibility/) · [56-second technical demo](https://youtu.be/wXoA9htThAM)
smithery -
Automate Cheat Engine using MCP protocol for use with AI!
MIT ★ 101 -
MCP server exposing game memory as 10 AI-callable tools (read/write player attributes, locate objects by value, scan memory)
MIT -
SinoutX (personal edition) — self-hosted, single-user AI workspace with a built-in assistant that captures, remembers and acts across web, Telegram and Viber. Your data, your server, your AI keys (BYOK). Ships with SINOUT_EDITION=solo.
NOASSERTION -
sentinelgithub.com/milema-gif/sentinelupdated this year
Behavioral guardrails for Claude Code — stop your agent from editing before it reads
MIT -
Persistent agent identity for AI agents — a hand-edited identity layer, nightly-distilled experience, and lessons injected every session, shared across your whole fleet. Works with Hermes, OpenClaw, Claude Code, Pi, and opencode.
npm PolyForm-Noncommercial-1.0.0 -
🌟Zero-dependency MCP kernel for coding agents. Runs and tests speculative edits in RAM shadow workspaces before touching disk.
MIT ★ 1 -
Notes for iPhone and Mac that ChatGPT, Claude, Claude Code and Codex can read and edit. Feels like Apple Notes, imports it too. Open source, encrypted.
MIT ★ 34 -
Connect your AI client to a local project to edit code, run tests, manage dev servers, and drive a browser with a live dashboard.
MIT -
Open-source MCP server and hook service that keeps Claude Code and Codex engineering context current as the code changes, and holds edits whose evidence changed.
npm MIT ★ 12 -
Rootz Archive: a Claude Code plugin that keeps an exact copy of every conversation and every file Claude edits, on your own computer, searchable by your AI. Free for individual users. No account, works offline.
NOASSERTION ★ 4 -
See, edit and delete what your AI tools remember. Local tool that supports Claude Code, Codex, Cursor, Gemini CLI, Windsurf, Copilot and 30+ more in a light-weight app. Platforms include Mac, Windows, Linux and your phone.
Apache-2.0 ★ 1 -
Home Assistant app (add-on) that exposes one directory as an MCP server: Claude and other clients read, search and edit files, read PDFs and query SQLite read-only. Built for a Karpathy-style LLM wiki, works for any file tree.
MIT ★ 4 -
Your Obsidian vault as Claude's second brain: self-hosted MCP connector for claude.ai, Claude mobile/Desktop and Claude Code. Links & backlinks, tags, frontmatter queries, daily notes, templates, canvases, safe concurrent edits. Local-first Markdown, no RAG. Docker + Cloudflare tunnel, OAuth 2.1.
MIT ★ 5 -
MCP server that keeps your notes, decisions, projects and documents in one folder on your own server: Claude and other AI assistants read, search and edit it under per-folder write policies, with no delete. Reads PDF, Word, Excel and e-mails as text. Docker, Linux or Home Assistant.
MIT ★ 2 -
Signed, append-only decision records for AI coding agents: an MCP server and CLI that hand your team's decisions to Claude Code, Codex, Copilot CLI, Gemini CLI, OpenCode and other MCP hosts, hold the edits a rule governs where the host's hooks allow, and let anyone verify the record offline.
npm Apache-2.0 ★ 5 -
PMX Canvas is a shared board where you and your AI agents collect and present the data and evidence behind your work. Curate that material into context for your agent's next step, then bring the results back onto the board to review and build on. You both add and edit material on the same board, in real time. Pin context to focus input and output.
npm MIT ★ 31