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.
-
Zero-dependency agent-to-agent codebase wiki for LLMs and AI agents. Autonomously maintained via record-change and mark-green for token-efficient lookup of files, dependencies, health, and summaries — across tiny scripts to 50k+ monorepos. Optional MCP server with rich tools for agents.
pypi MIT -
Keep decision records in Git and see which ones a code change touches, in pull requests and through AI agents.
Apache-2.0 ★ 2 -
Check what your AI agents actually changed. AI agent safety for databases and health records: rules decide each tool call, a check compares what really changed with what the agent reported, and each action is sealed as a signed record anyone can check offline. Run npx deedglass and a simple local app opens in your browser. Apache-2.0 core.
go Apache-2.0 ★ 7 -
Read-only decision record: query decisions, drain decision changes, verify token.
npm MIT -
MCP server and CLI for b4-after: lets coding agents model a system as C4 (LikeC4), propose and review changes to it, promote accepted changes, and record the evidence that verifies them
npm MIT -
Local MCP server that turns screen recordings and video URLs into keyframes, OCR, audio transcripts, and timelines for Claude Code. Watch a YouTube tutorial, search past sessions, long-poll for screen changes.
MIT -
Read your recorded AI agent runs, and what changed when you swapped a model or prompt. Read-only MCP server for Pramana.
pypi Apache-2.0 -
Chrome bookmarks as MCP tools, with every change recorded in git so it can be rolled back
npm MIT -
Local flight recorder for macOS: records app launches, sleep/wake, memory, CPU, battery, network, disk and file changes, and explains them with cited evidence and stated confidence. CLI, menu bar app and a read-only MCP server.
MIT ★ 2 -
MicrosoftGraph_Transcript_Webhook_MCPgithub.com/ITSpecialist111/MicrosoftGraph_Transcript_Webhook_MCPupdated this year
An MCP server for managing Microsoft Graph webhook subscriptions for Teams meeting transcript and recording change notifications. Uses **application-level** (client credentials) authentication.
★ 3 -
Turns any text stream into a dated history of how its subject changed. Point it at your Claude Code, OpenCode or Codex sessions to track how you think and work, or at market news, research, or any corpus to track how it changed — raw records to observations to bi-temporal facets to narrative, with provenance, served over MCP. Single pure-Go binary.
npm MIT ★ 12 -
Git records what changed. AgentCommons records why. Design decisions, tradeoffs and dead ends from every Claude Code, Codex, Gemini CLI and pi session, searchable by your team and its agents.
Apache-2.0 ★ 1 -
MCP server for managing Cloudflare DNS records. Use natural language with your favourite AI tool e.g Claude Desktop or Claude Code to list domains, view DNS records, and make DNS changes.
npm MIT ★ 2 -
IoT-focused pure-Rust Wi-Fi diagnostics library with change-only recording and typed verdicts
Apache-2.0 ★ 1 -
trialwatchgithub.com/cognis-digital/trialwatchupdated this year
Query, diff, and monitor ClinicalTrials.gov records, alerting on status, enrollment, or result changes.
NOASSERTION ★ 1 -
Get WordPress work done from your AI assistant. Preview changes, keep a history, and record what future edits must preserve.
oci MIT -
Your agent said done. It never watched it run. Let AI agents see UI recordings: changed regions, flicker, frozen motion. CLI + MCP + skill.
npm MIT -
Multi-model review gate for AI-written code: independent models from different providers cross-check every change, and a deterministic gate records PASS, BLOCKED or INCONCLUSIVE. CLI + MCP server. MIT.
MIT -
Flight recorder for AI agents: append-only event log of tool calls, errors, handoffs, file changes and token costs, with a query API, dashboard, and Claude Code hooks.
MIT -
kryptheon-mcpactive
Kryptheon for AI coding tools: record a flow, check it after every change, and attack a copy of the database - from Cursor, Claude Code, Codex, Windsurf or Antigravity.
npm MIT -
Siljangnim for MCP agents: a creative coding studio your agent drives — make live visuals by talking, change and mix them, perform with MIDI/OSC/audio, record video
npm GPL-3.0-or-later -
MCP server for the Vox Trust protocol: verify sealed voice recordings (WAV) from any AI assistant, with the seconds that changed. Verify-only, no secrets. Pre-1.0, not audited.
npm Apache-2.0 -
Salesforce unlocked package that records which external systems depend on which objects and fields, so you can check impact before changing metadata. Includes LWC dashboard, Flow/Agentforce action and Claude via Hosted MCP.
MIT -
e2easy-plugingithub.com/e2easy-app/e2easy-pluginupdated this year
Keep E2Easy recorded tests in sync with UI changes. Pre-commit plugin for Cursor and Claude Code - MCP sync, stable test attributes, and local Playwright validation.
MIT -
An undo layer for AI agents. Records what an agent changed through MCP, puts it back, and refuses to let it do the unforgivable unasked.
npm MIT ★ 3 -
An MCP server that gives a coding agent a record it did not write: what actually changed on disk and in git, held against what the agent said it did.
npm MIT -
A knowledge graph engine for codebases that : recorded intent, co-change context, ranked blast radius, architecture context, precise code discovery, continuous history, visible context accounting and freshness awareness.
Apache-2.0 ★ 9 -
See what your Node backend really does on every request, catch what a change broke across every route, and prove fixes by replay. A local recorder, dashboard and MCP server, with a stop hook for coding agents.
npm MIT -
Small Print (MCP): the GitHub Action that fails a build when the small print your agents read has changed, the smallprint command, and the MCP server that reads the public record. smallprint.dev
npm MIT ★ 1 -
MCP server for Proof: an audit trail for AI-written code. Records what the agent changed, what a person actually reviewed, and who signed off. Works inside Claude Code, Claude Desktop and Cursor.
npm AGPL-3.0 -
Span is a local MCP server that keeps the semantic understanding your coding agent worked out — the conclusions, the exact evidence behind them, and what needs another look when that evidence changes. Three tools: ask, record, reconsider.
pypi Apache-2.0 ★ 1 -
Swarm Verify: independent verification for software changes from AI coding agents or humans. Runs a project's real checks, separates what passed from what was established, names what stayed unmeasured, and records evidence anyone can check. CLI, GitHub Action, Claude Code hook, MCP server, pre-commit hook.
ISC ★ 2 -
Genealogy research toolkit for AI assistants (MCP server): 179 archives and databases (Russia and the former USSR, Europe, Latin America, US), where a province's records are kept today, live search, name variants and Daitch–Mokotoff, Julian/Hebrew/Republican calendars, church feasts, kinship and DNA, evidence checks, a research project that remembers, your own GEDCOM tree read and changed in place without losing anything, GEDCOM import/export.
pypi MIT -
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 -
Vetono update data
93 agentic MCP tools + 49 specialist agents for every major AI CLI — self-learning, no API keys. Veto gives Claude Code, Codex CLI, Gemini CLI, Cursor, Windsurf, Zed, and JetBrains a council of specialist agents, persistent cross-platform memory, a self-learning router, CI/CD gates, and dependency-hallucination + decision-drift guards — all powered by your existing AI subscription. No separate API keys, no extra billing. **Highlights** - **Council debate** — multi-agent GREEN/YELLOW/RED verdicts on risky changes (LLM-backed via MCP Sampling, with deterministic fallback) - **Cross-platform memory & session handoff** — save on Claude, resume on Gemini or Codex; nothing needs re-explaining - **Self-learning router** — improves task routing from recorded outcomes - **Guards & gates** — CI gate, diff review, secrets scan, dependency-hallucination verification, decision-drift enforcement - **Compact mode** — advertise 9 tools instead of 93 and save ~16K context tokens; every tool stays callable **Links**: [GitHub](https://github.com/jigyasudham/veto) · [npm](https://www.npmjs.com/package/@jigyasudham/veto)
smithery -
Telegram AI CLI MCPgithub.com/stufently/telegram-ai-cli-mcpno update data
A CLI and MCP server, built from one Python codebase, for a personal Telegram account over MTProto (not the Bot API). Before the first start, sign in from a terminal with tg-ai account login; the server has no login flow of its own and will not work until that is done. Read tools cover chats, inbox, message search, who a username belongs to, group members, folders, drafts, scheduled messages, mentions, sessions and a waiting watch; an optional local archive answers searches offline. Anything that changes Telegram (sending, replying, joining, deleting, moderation) is a typed plan tool, available only under the plan profile: it records the intention, and a person applies it with tg-ai plan apply. No tool applies a plan, and reading never marks a chat read. Direct messages stay closed until allowlisted. Telegram may limit accounts it judges to be running abusive automation; use on your own account is at your own risk.
smithery -
Bernsteinno update data
Deterministic, verifiable orchestration for CLI coding agents (Claude Code, Codex, Gemini CLI, Aider, Cursor and 40+ more), exposed over MCP. Bernstein runs orchestration tasks in isolated git worktrees, records a lineage and replay journal for every run, and can anchor task claims, progress updates and posted artifacts in an opt-in HMAC-chained audit log. Tools: 14 are advertised at the default `standard` tool tier. Set BERNSTEIN_MCP_TOOL_TIER to change the tier: `core` advertises 5, `standard` advertises 14 (default), `all` advertises 18. Transports: stdio (default), Server-Sent Events, and streamable HTTP. Install and run: uvx --from bernstein bernstein mcp or pip install bernstein && bernstein mcp HTTP transport: bernstein mcp --transport http Authentication: the stdio transport needs none. The HTTP transport accepts a static bearer token via BERNSTEIN_MCP_TOKEN or BERNSTEIN_AUTH_TOKEN, and binds to loopback only when neither is set. Documentation: https://bernstein.readthedocs.io Requires Python 3.12 or newer. Package `bernstein` on PyPI, version 3.9.0.
smithery -
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 -
Shows what your iOS app is actually doing — the screen, the app's own state, its network traffic and its logs — to an AI coding agent on your Mac, so checking a UI change stops being a manual tap-and-screenshot loop.
MIT -
decreegithub.com/doruksahin/decreeupdated this year
Decision governance for code and the agents editing it — track PRD→ADR→SPEC decisions, map them to files, and gate changes against them. Deterministic, honest CLI + MCP.
MIT ★ 2 -
⛏️ Mole in your editor + agent: team decisions beside the code, checked as you change it. VS Code extension, Claude Code plugin, MCP, CLI.
MIT -
Your AI agents know the code. They don't know the company. Free, local CLI and MCP server that imports your team's decisions from Jira, Confluence, Linear, Notion, Slack, Teams, Zoom, GitHub, GitLab, git history and repo docs, and checks your coding agent's change against them before it edits.
npm MIT ★ 14 -
Review your AI agent's edits like tracked changes: marked in VS Code and the Markdown preview until you accept them.
MIT -
Building your AI organization: shared records and handoffs, plus a merge gate for AI-written code with signed, replayable attestation. Works with Claude Code, Codex, Cursor, and Gemini CLI.
npm MIT ★ 23 -
Find out why a component exists and what changes if a decision went the other way. Reads the ADRs, RFCs and design notes already in your repo, cites the passage behind every inferred claim, keeps human review on top, and previews option switches on a map. Local-first, Markdown only, MCP tools for agents.
Apache-2.0