git-history-viewer
A local-first MCP server that provides read-only access to Git repository history, exposing commit metadata and optionally content patches via configurable policies, with optional AI-powered analysis while never modifying Git state.
README
Git History Viewer
A free, local-first, read-only Git history graph for the browser, CLI, and MCP clients.
Public alpha. Pin an exact version in managed environments.
Install
Requires Node.js 22 or later and Git 2.40 or later.
npm install --global @peekaboo-develop/git-history-viewer@alpha
Open a repository
cd /path/to/repository
git-history-viewer web .
The viewer opens a loopback-only browser session for that working copy. It does not clone, fetch, pull, push, or modify Git state. Remote-tracking refs show the last state fetched by your existing Git workflow.
For MCP setup, start with metadata-only access:
git-history-viewer mcp --repo /absolute/path/to/repository --content-policy metadata
Development
npm install
npm test
npm run license:check
npm run test:pack
node bin/git-history-viewer.mjs web . --port 0 --no-open
node bin/git-history-viewer.mjs web . --port 0 --no-open --ollama-model qwen3:4b
node bin/git-history-viewer.mjs config path
node bin/git-history-viewer.mjs config validate
node bin/git-history-viewer.mjs profiles list
node bin/git-history-viewer.mjs web . --ai-profile local-fast --ai-profile local-deep
node bin/git-history-viewer.mjs mcp --repo /absolute/path --content-policy metadata
node bin/git-history-viewer.mjs cache status
The default MCP policy does not expose patches, author emails, or absolute paths. The application never fetches or mutates Git state.
The Web header includes an AI接続 guide for verified Codex, Claude Code, and Cursor MCP setup templates. It never edits a client configuration or exposes the repository's absolute path. A selected commit can produce a short MCP prompt containing only its OID.
The optional Web AI flow is disabled unless an --ai-profile or the legacy --ollama-model option is supplied. Profiles support numeric-loopback Ollama and the official OpenAI, Anthropic, and Google APIs. Every provider previews the exact metadata payload and requires a second click before generation. When multiple profiles are enabled, the detail pane provides a model selector. Remote endpoints and credential names are fixed in code: OpenAI reads OPENAI_API_KEY, Anthropic reads ANTHROPIC_API_KEY, and Google reads GEMINI_API_KEY. Keys are never accepted from the browser or config file. Loopback does not by itself prove where Ollama executes a model. Patches, file contents, identities, absolute paths, refs, remotes, and commit IDs are not sent.
The config path is OS-specific and printed by config path. The strict JSON format stores no credentials, arbitrary endpoints, prompts, or environment-variable names:
{
"schemaVersion": "1",
"defaultProfileId": "local-fast",
"profiles": [
{
"id": "local-fast",
"label": "Local fast",
"provider": "ollama",
"model": "qwen3:4b",
"ollamaPort": 11434,
"maxOutputTokens": 1536
},
{
"id": "openai-fast",
"label": "OpenAI fast",
"provider": "openai",
"model": "gpt-5.4-mini",
"maxOutputTokens": 1536
},
{
"id": "claude-balanced",
"label": "Claude balanced",
"provider": "anthropic",
"model": "claude-sonnet-4-6",
"maxOutputTokens": 1536
},
{
"id": "gemini-fast",
"label": "Gemini fast",
"provider": "google",
"model": "gemini-2.5-flash",
"maxOutputTokens": 1536
}
]
}
Config changes require a restart. --ai-profile is repeatable and is the explicit allowlist for that Web session. Export the corresponding fixed key only when selecting a remote profile.
The AI-result cache uses bounded, sharded JSON files in the operating system's user cache directory. Cache metadata stores no API keys, raw prompts, diffs, repository paths, or remote URLs; generated results may still repeat sensitive source text. --no-ai-cache bypasses reads and writes. cache clear removes only recognized cache-entry files.
The commit detail pane can recommend up to two version-neutral links from a compiled registry of official documentation for strong path markers such as GitHub Actions workflows, Docker/Compose files, TSConfig, Vite config, and Vue SFCs. Link display performs no network access. A separate button can fetch bounded excerpts directly from those fixed official URLs through a DNS-rebinding-resistant, pinned-IP HTTPS transport. The destination can infer the detected technology, but repository names, paths, commits, messages, cookies, credentials, and referrers are not sent. Redirects are rejected. When AI is enabled, another separate preview and confirmation can send the displayed excerpts plus opaque citation IDs to the selected provider; official URLs and titles are never sent to the model and are mapped back only by the server. Version detection remains unavailable.
See docs/PRODUCT_SPEC.md, docs/ARCHITECTURE.md, docs/MCP.md, docs/SECURITY.md, and docs/ROADMAP.md for the normative design and planned remote-environment support.
Release preparation and rollback are documented in docs/RELEASING.md. Building an artifact does not authorize publishing it.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.