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.
-
@psno/subconductorgithub.com/paulbenchea/mcp-subconductorupdated this year
Subconductor is an MCP server that acts as a Persistent State Machine for AI agents. It prevents 'context drift' by externalizing the task list into a Markdown table manifest, supporting batch operations, dynamic custom columns, task IDs, and non-linear e
npm Apache-2.0 -
Durable, policy-scoped semantic context for AI agents via MCP: source authority, state reconciliation, SQLite persistence, and local demos.
MIT -
Persistent, branch-aware workflow state memory MCP server for AI coding assistants. Tracks tasks, accepted decisions, and active blockers to prevent session context bloat and speed up development.
npm MIT ★ 45 -
Persistent multi-agent work graph and document-state machine for contracts, specs, slices, evidence, verification, handoffs, and cost-aware AI execution.
Apache-2.0 -
Memory that forgets: a persistent, self-revising belief-state MCP server for agentic R&D. Uses a hypothesis DAG to auto-prune dead ends and deduce truth without extra probes.
pypi MIT -
Gravitas-Coregithub.com/ahmed-coding/Gravitas-Coreupdated this year
Gravitas-Core is a Python-native autonomous AI control plane and MCP (Model Context Protocol) server. It provides a persistent execution core for intelligent agents with task state management, memory persistence, terminal/browser engines, and recursive project intelligence—all designed for scalable AI-driven workflows.
MIT ★ 1 -
Persistent, stateful remote SSH sessions for AI agents via MCP. Keep cwd, env & shell state across commands — powered by your local OpenSSH. Works with Claude, Cursor & any MCP host.
npm MIT ★ 1 -
Stememgithub.com/JOJO-Adam/Stemem#readmeno update data
Identity-layer MCP runtime for AI agents: persistent personality state, emotion-driven tone drift, compression-immune memory. Local-first, zero egress, zero dependencies.
smithery -
nexusgithub.com/siva-nagendra/nexusupdated this year
In-game MCP server for Unreal Engine with autonomous AI agents, native scene state awareness, and persistent learning.
-
Handyman-MCPgithub.com/kennetchau/Handyman-MCPupdated this year
Stateful AI Agent Infrastructure with Persistent FTS5 Memory and Sandboxed Execution.
MIT ★ 1 -
A multi-database MCP server with explicit persistent sessions, continuous SQL state, transaction tools, and guarded writes.
npm MIT -
Remote MCP server for persistent developer knowledge — sprint docs, architecture decisions, project state. Deploy once to Azure, share one brain across Claude, Copilot, and Cursor.
MIT ★ 1 -
MCP server that dispatches and supervises parallel Codex CLI workers, one isolated Git worktree each, with a persisted work state machine.
npm MIT -
Persistent memory layer for AI coding agents — current-state retrieval that knows what is still true. Claude Code, Codex, Gemini CLI.
Apache-2.0 ★ 11 -
claude-book-agentgithub.com/raghunandanjp-dotcom/claude-book-agentupdated this year
AI-powered MCP server for Claude Desktop that automates structured children's book generation through workflow orchestration and persistent state management.
-
MCP server that drives a real Chrome browser via CDP + Playwright: structured snapshots, click/type, screenshots, network inspection. Persistent login state in an isolated profile.
NOASSERTION -
pagerunnergithub.com/enreign/pagerunnerupdated this year
Persistent browser layer for AI agents. Connects any MCP client to your real Chrome profiles — already logged in, with site intelligence that compounds over time. Sessions survive daemon restarts. PII anonymization, encrypted state, auto-checkpoints, macO
npm MIT -
engramgithub.com/HBarefoot/engramupdated this year
🧠 Persistent memory for AI agents. SQLite for agent state. Zero cloud dependencies. Local embeddings. MCP-native integration with Claude Desktop/Code, Cursor, Windsurf & more.
MIT ★ 9 -
MCP server that acts as a persistent context engine for AI assistants: task, epic and dependency tracking in SQLite/PostgreSQL so agents keep perfect state across tool calls and sessions
MIT -
Memory that forgets: a persistent, self-revising belief-state MCP server and Python API for agentic R&D. Uses a hypothesis DAG to auto-prune dead ends and deduce truth without extra probes.
MIT ★ 12 -
Secure, persistent, and cross-session agentic memory for commerce state. This skill enables OpenClaw nodes to maintain granular historical context of vendor negotiations, discount eligibility, and user preferences over long-term fulfillment cycles.
MIT -
Universal Git-powered persistent memory, context rules & native MCP server for AI coding agents. Prevents context amnesia with stateful workflows, Git Notes, and cross-model review gates across Claude, OpenAI, Gemini & DeepSeek in Cursor, Claude Code & VS Code.
pypi MIT ★ 6 -
Stateful SageMath MCP server: 40 tools for symbolic math, algebra, number theory, graphs, plotting and claim verification. A persistent Sage session per client, an AST policy in front of it, and a hardened container as the real boundary.
MIT ★ 19 -
Model Context Protocol (MCP) server backed by CloakBrowser stealth Chromium. Drop-in replacement for @playwright/mcp with source-level fingerprint patches, humanized inputs, and persistent profiles. 86 tools across navigation, interaction, page state, net
npm MIT -
Execute terminal commands locally or remotely via SSH with session persistence and environment variable support. Manage terminal sessions that maintain state for up to 20 minutes, enabling efficient command execution workflows. Connect using stdio or SSE for flexible integration with AI models and applications.
MIT ★ 2 -
claudexgithub.com/unlikesooraj/claudexupdated this year
Connect Claude Code with Codex CLI — seamless local session bridge for sharing context across Claude Code and OpenAI Codex CLI without losing any state. Smart cross-tool session sharing, persistent rolling context, built-in MCP server, no API keys, no cloud.
MIT -
CodeBoxMCPgithub.com/MadhanMohanReddy2301/CodeBoxMCPupdated this year
Code Box is a self-hosted, open-source Model Context Protocol (MCP) server that provides AI agents with a secure, stateful environment to execute Python and JavaScript code. It acts as a direct, zero-cost replacement for Azure Assistants Code Interpreter by persisting variables across calls, and safely executing read-only SQL querys
MIT -
MCP server for Claude Code, powered by the v4 semantic runtime: builds persistent, budget-aware repository intelligence from AST/graphs, serves bounded context packages, preserves agent state across /compact, and provides end-to-end edit impact analysis in a single Rust binary.
-
Stop AI coding agents from guessing. MCP server + Claude Code plugin that enforces intent alignment BEFORE coding: PRD → intent-confidence gate → Mermaid contracts (state machines / sequence diagrams / decision tables) → mechanical lint with zero CRITICAL → then code. Human-in-the-loop, file-persisted, zero-config.
MIT ★ 3 -
ab-variant-agentgithub.com/beautifulgownss/ab-variant-agentupdated this year
A LangGraph-powered agent that generates A/B copy variants for marketing campaigns (email, SMS, landing pages) using distinct persuasion angles. Features structured LLM outputs, validation with retries, Postgres checkpointing for state persistence, and an MCP server for tool-based integration. Built as a foundation for human-in-the-loop review.
-
mcp-subconductorgithub.com/PaulBenchea/mcp-subconductorupdated this year
Subconductor is an MCP server that acts as a Persistent State Machine for AI agents. It prevents "context drift" during long refactors by externalizing the task list into a physical manifest. It enables non-linear execution, allowing agents to handle dependencies dynamically while maintaining a transparent, resumable audit trail.
Apache-2.0 ★ 2 -
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 -
Persistent project workflow for Codex that connects intent, sources, plans, tools, evidence, and continuity, with a read-only Windows Studio.
MIT