Depverse

📖Full documentation →Install, tool reference, and setup guides for Claude Code, Claude Desktop, Cursor, Cline, Windsurf, and Copilot.

Depverseis anMCP(Model Context Protocol) server that exposes the publicnpm Registryas 39 structured tools that Claude (or any MCP-aware client) can call — versions, dependencies, changelogs, security advisories, download trends, and more — without ever leaving the editor.

Depverse/ ├── mcp_server.py # The MCP server — all 39 npm tools live here ├── mcp_client.py # Thin MCP client wrapper (stdio transport) ├── test_npm_tool.py # Manual end-to-end test for the server ├── pyproject.toml ├── uv.lock ├── .mcp.json # Example MCP server config for external clients ├── docs/ # Documentation site (GitHub Pages) │ │ # --- Optional: bundled CLI chat (main.py) --- ├── main.py # Entrypoint for the optional CLI chat └── core/ ├── chat.py # Tool-using chat loop ├── cli_chat.py # CLI-flavoured chat (supports @docs and /commands) ├── cli.py # prompt-toolkit UI (autocompletion, history, key bindings) ├── claude.py # Anthropic API wrapper └── tools.py # Bridges MCP tool calls into Anthropic tool_use blocks

- Python3.10+
- uv(recommended) or plainpip
- Network access toregistry.npmjs.org,api.osv.dev,api.npmjs.organd (for changelogs)api.github.com/raw.githubusercontent.com

git clone https://github.com/yash-neural/Depverse.git cd Depverse uv venv source .venv/bin/activate # Windows: .venv\Scripts\activate uv pip install -e .

python -m venv .venv source .venv/bin/activate pip install -e .

uv run test_npm_tool.py # spawns the server, lists tools, calls each once

{ "mcpServers": { "Depverse": { "type": "stdio", "command": "uv", "args": [ "--directory", "/absolute/path/to/Depverse", "run", "mcp_server.py" ], "env": {} } } }

Use"command": "python"and drop"run"fromargsif you aren't usinguv.

test_npm_tool.pyspawns the server, lists its tools, and calls each one with a realistic input. Handy for verifying changes without the full Claude loop:

It prints each tool call and the (truncated) JSON response, so you can eyeball the output.
- mcp_server.pyregisters every tool withFastMCP(from the MCP Python SDK). Each tool is anasyncfunction that talks to the npm Registry, OSV.dev, or the npm download API overhttpx, with a shared_fetch_jsonhelper that enforces a 10 s timeout and consistent 404 / error messaging.
- mcp_client.pywrapsmcp.ClientSessionwith a small context-managed class — only used bytest_npm_tool.pyand the optional CLI chat.
- The MCP client (Claude Code, Claude Desktop, Cursor, etc.) spawnsmcp_server.pyas a subprocess over stdio. JSON-RPC frames flow in both directions; tool calls return structured JSON the model can reason about.

The MCP server is the whole point — everything incore/is scaffolding for theoptionalbundled CLI chat, which you can ignore if you're just plugging Depverse into Claude Code.

No license file is committed yet — add one (MIT is a sensible default) before publishing or accepting external contributions.

Make git commits on behalf of AI to track AI contributions in your codebase.

AI drafts the commits, you decide what ships. GitInProgress lets your AI read your uncommitted working tree and propose how to split it into focused commit candidates — each with a draft commit message. Nothing lands in git history until you review and approve; the AI never commits on its own.

An MCP server for interacting with and automating Git repositories using Large Language Models.

Generates Pull Request messages based on Git commits and changes.

Packs repository contents into a single file for AI consumption.

Entity-level code intelligence for agents

Local stdio MCP server that lets AI coding agents read and maintain structured architecture, rules, and decisions directly from your repository.

Anchor is local repo and org memory for AI coding agents. It indexes GitHub PR history, current code, tests, regressions, architecture, and cross-repo impact locally, then exposes concise cited context through MCP and CLI workflows. Local-first. Read-only GitHub access. No CLI telemetry. No SaaS. No remote LLM calls.

An intelligent system for managing programming rules, supporting search, versioning, code validation, and prompt enhancement.

DOS is a small open-source kernel that doesn't believe your AI agents: it verifies what actually shipped from git evidence, arbitrates file collisions between parallel agents, and refuses with structured reasons.