Agent Guidance MCP

MCP server serving AI agent guidance through a 168-skill catalog, bundled guidance corpus, workflow prompts, and bounded project-code context tools over Stdio transport.

Skills are sourced from Everything Claude Code (ECC) v2.0.0 and community contributions, covering backend, frontend, testing, security, DevOps, data, research, and 12+ language ecosystems.

Installation

Install the Agent Guidance MCP server and configure all local IDE clients with a single command:

Linux / macOS (Bash):

curl -fsSL https://raw.githubusercontent.com/JunMystery/Agent-Guidance-MCP/main/scripts/install.sh | bash

Windows (CMD / PowerShell):

powershell -Command "irm https://raw.githubusercontent.com/JunMystery/Agent-Guidance-MCP/main/scripts/install.ps1 | iex"

This requires no prior Python installation; the script will automatically bootstrap uv (a single-binary Python toolchain) to run the server if Python is not present.

Upgrading

To update the server and refresh your IDE registrations: Simply rerun the installation command.

To update the standards catalog & skills only (one line):

agent-guidance-mcp --update

To update the executable code package only:

uv tool update agent-guidance-mcp

Uninstalling

Linux / macOS (Bash):

curl -fsSL https://raw.githubusercontent.com/JunMystery/Agent-Guidance-MCP/main/scripts/uninstall.sh | bash

Windows (CMD / PowerShell):

powershell -Command "irm https://raw.githubusercontent.com/JunMystery/Agent-Guidance-MCP/main/scripts/uninstall.ps1 | iex"

This will automatically remove the server registration from all detected IDE client configurations, clean up the global rules in AGENTS.md, remove the database/skills directory (~/.agent-guidance), and uninstall the executable.

Manual / Local Developer Install

python -m venv .venv
.venv/bin/pip install -e ".[dev]"      # Linux / macOS
.venv\Scripts\pip install -e ".[dev]"  # Windows

Run the server module:

agent-guidance-mcp
.venv/bin/python -m agent_guidance_mcp          # Linux / macOS
.venv\Scripts\python.exe -m agent_guidance_mcp  # Windows

To point the server to a different standards corpus, set:

AGENT_GUIDANCE_ROOT=/path/to/Agent-Guidance

Platform notes and client-specific setup are covered in Installation and Client Configuration.

Quick Start

Run the server through an MCP client, or verify it locally with MCP Inspector:

DANGEROUSLY_OMIT_AUTH=true npx @modelcontextprotocol/inspector .venv/bin/python -m agent_guidance_mcp

Then call task_pipeline(...) to load task guidance and bounded project context before editing. See Usage Guide for practical workflows.

Documentation

• Installation - automatic and manual setup.
• Client Configuration - VS Code, GitHub Copilot, Claude Desktop, Cursor, Gemini-compatible config, and env vars.
• Usage Guide - quick checks, recommended agent workflows, and examples.
• MCP Surface Reference - all tools, prompts, and resources.
• Project Context Tools - grouped tree, search, file read, snapshot export, token guidance, and freshness rules.
• Development Guide - tests, project structure, and maintainer notes.
• Repo Map For Agents - existing repository orientation notes.
• MCP Integrations Guide - SQLite caching, CodeGraph-like AST parsing, and Context7 docs details.

MCP Surface

Tools:

• task_pipeline(task, project_path, focus, code_query, include_tree, include_ui, limit) — Recommended first call. Prepares task recommendations, project context, and optional UI guidance.
• guidance(operation, query, identifier, category, kind, limit, include_content) — Standards catalog operations: list, get, search, recommend, reason, and docs (live library docs via Context7).
• project_context(operation, project_path, query, relative_path, start_line, max_lines, max_depth, output_path, max_file_bytes, max_total_bytes, limit) — Bounded project context: tree, search, read, snapshot, symbols, references, structure, callers, and callees (local SQLite CodeGraph engine).
• ui_ux(operation, query, domain, stack, project_name, output_format, limit) — UI/UX Pro Max: search, design system, slides.
• health_check() — Server health status and metadata.
• token_stats() — Token optimization statistics for the session.

Resources:

• standards://manifest — Indexed standards manifest (JSON)
• standards://skill/{name} — On-demand skill capsule (Markdown)
• standards://document/{identifier} — Standards document by slug (Markdown)
• standards://version — Server version info (JSON)

Prompt:

• workflow_prompt(mode, subject, target) — Load a workflow prompt by mode (plan, test, deploy, debug, etc.)

Development

python -m pytest

The test suite verifies catalog discovery, MCP handler registration, standards search, recommendation behavior, and project-context tooling. See Development Guide for more detail.