Stackguide Mcp

io.github.isucorp-taimiralainPublished by GitHub account isucorp-taimiralainv4.1.3Updated Oct 24, 2026

Dynamic context loading for AI assistants with TDD workflow and AST analysis

Overview

AI-generated overview

StackGuide loads project-aware rules and context for AI coding assistants and runs an active TDD workflow with planning, verification, and release steps.

What you can do
StackGuide is a local MCP server for AI coding assistants in Cursor, VS Code, JetBrains, and Visual Studio. It detects the project stack, loads architecture and methodology rules, scores code health, generates boilerplate, and runs an execution-first agent workflow with intake, plan, verify, and release actions. It also connects to trackers such as GitHub, GitLab, and Jira, and can create Jira tickets from a strict MAIN DESCRIPTION template.
When to use it
Use it when a team wants consistent implementation quality and a repeatable TDD-style delivery flow inside the IDE, including ticket intake, planning, test and lint verification, and release checks. It is less useful for one-off coding questions or projects that do not want a shared workflow configuration committed to the repository.
Requirements
Runs locally over stdio, installed via the npm package @stackguide/mcp-server, typically launched with npx. Requires Node.js and a desktop MCP client such as Cursor, VS Code with Copilot Chat, JetBrains, or Visual Studio. No authentication is declared. Tracker integrations such as Jira, GitHub, and GitLab require their own credentials, and the optional STACKGUIDE_INTEGRITY_KEY environment variable can be set for stronger tamper resistance.
Before you install
The server writes files into the repository, including stackguide.config.json, which is meant to be committed, and local .stackguide artifacts such as analysis-cache.json and health-history.json, which must not be committed. The agent workflow can create Jira tickets, tags, and pull or merge requests, and runs tests, lint, and build commands. It checks npm once a day for updates unless STACKGUIDE_NO_UPDATE_CHECK=1 is set. Tracker and VCS integrations need credentials, and…

Installation

In SourceWeft

  1. Open Stackguide Mcp in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

StackGuide MCP Server

Dynamic context + active engineering workflow for AI coding assistants in Cursor, VS Code, and other MCP clients.

[npm version] [license] [mcp registry]

What It Does

@stackguide/mcp-server helps teams standardize implementation quality with:

  • project-aware rule/context loading
  • code review and health scoring
  • auto-detected setup and stack scaffolding
  • active agent workflow that executes real work (intake, plan, verify, release)
  • adaptive TDD preferences per project (model + token profile)
  • Jira ticket creation from strict MAIN DESCRIPTION templates
  • tracker + VCS + test orchestration for delivery flow

This server is built for real team usage, not just prompt templates.

What's New In v4.x

Full history: CHANGELOG.md.

  • v4.3.1 - Update notifier: StackGuide checks npm once a day in the background and tells you once per session when a newer version exists, with a safe update command. Disable with STACKGUIDE_NO_UPDATE_CHECK=1.
  • v4.3.0 - Team mode:
    • setup wizard:true asks stack, architecture, methodology, conventions and tracker, then writes a shared stackguide.config.json at the repo root. Commit it so every teammate gets the same context.
    • On startup StackGuide finds that file (also from subfolders) and loads it without running setup or re-detecting the project. context returns one guidance line per architecture and methodology.
    • Optional ruleCategories limits which rules are loaded (saves tokens). Existing healthWeights in the file are preserved.
    • Multi-agent roles that follow the IDE you are using (detected from the MCP client). Defaults: Claude Code opus / haiku / haiku; Cursor inherit (your selected model, e.g. Composer or Grok) / fast / fast; other IDEs inherit. The wizard asks which model each role should use (recommended default, your most used models first, or type any id your IDE lists); answer with models:{planner,implementer,verifier} or models:{} to accept the defaults. Choices are saved per IDE in stackguide.config.json, so mixed teams do not clash.
    • agent action:"plan" returns a delegation block for the current IDE, and the wizard writes stackguide-implementer.md / stackguide-verifier.md into .claude/agents/ and .cursor/agents/. Edited files are kept; only their model: line is refreshed.
    • Re-running setup wizard:true edits the saved config; .stackguide/ stays local and guarded.
  • Added active agent tool (execution-first workflow).
  • init now generates .stackguide/config.json automatically.
  • Added tracker service support for GitHub, GitLab, and Jira.
  • Added VCS service for branch checks, commit parsing, CI status, and PR/MR creation.
  • Added test runner service for structured test/lint/build execution.
  • Updated TDD prompts to use active tool calls instead of large passive markdown payloads.
  • Hardened local persistence:
    • .stackguide artifacts removed from version control
    • repository guard blocks accidental tracking of local artifacts
    • cache and health history include integrity validation and safer writes

Install By IDE

Cursor

  • File (workspace): .cursor/mcp.json
  • File (global): ~/.cursor/mcp.json
json
{  "mcpServers": {    "stackguide": {      "command": "npx",      "args": ["-y", "@stackguide/mcp-server@latest"]    }  }}

JetBrains (IntelliJ, WebStorm, PhpStorm, etc.)

  • Open Settings | Tools | AI Assistant | Model Context Protocol (MCP).
  • Click Add and paste JSON config.
  • Recommended: set it as Project-level unless you want global scope.
json
{  "mcpServers": {    "stackguide": {      "command": "npx",      "args": ["-y", "@stackguide/mcp-server@latest"]    }  }}

VS Code

  • Requirement: GitHub Copilot Chat with MCP enabled.
  • File (workspace): .vscode/mcp.json
  • VS Code MCP schema uses servers as root key.
json
{  "servers": {    "stackguide": {      "type": "stdio",      "command": "npx",      "args": ["-y", "@stackguide/mcp-server@latest"]    }  }}

Visual Studio (Windows, 2022 17.14+ / 2026+)

  • Enable GitHub Copilot Agent mode.
  • Supported config locations include:
    • %USERPROFILE%\\.mcp.json (global)
    • <solution>\\.mcp.json (repo-scoped)
    • <solution>\\.vs\\mcp.json (solution/user-scoped)
  • Visual Studio MCP schema uses servers as root key.
json
{  "servers": {    "stackguide": {      "type": "stdio",      "command": "npx",      "args": ["-y", "@stackguide/mcp-server@latest"]    }  }}

How To Use (Any IDE)

  1. Restart the IDE after adding the MCP configuration.
  2. Open AI chat in agent/tool mode.
  3. Call setup to detect project context.
  4. Call init action:"full" to scaffold .stackguide/ and defaults.
  5. Use agent actions (intake, create_ticket, plan, verify, release) for delivery flow.

Quick Start

Team setup (once per project, then commit stackguide.config.json):

bash
setup wizard:truesetup wizard:true type:"nextjs" architecture:"clean" methodology:"tdd" conventions:["Use named exports"] tracker:"jira"

Personal setup:

  1. Configure your project:
bash
setup type:"react-typescript" enableAdaptiveTdd:true model:"gpt-5" integrations:["jira","github"] tokenMode:"compact"init action:"full" model:"gpt-5" tokenMode:"compact" integrations:["jira","github"] mcpSyncTargets:["cursor","root"] applyMcpTemplates:true
  1. Check generated setup:
bash
init action:"status"agent action:"status"
  1. Run active TDD workflow:
bash
agent action:"intake" ticket:"PROJ-123"agent action:"create_ticket" mainDescription:"<MAIN DESCRIPTION>" projectKey:"PROJ"agent action:"plan" brief:"<brief-from-intake>"agent action:"verify"agent action:"release" version:"v1.2.0"

Tools (16)

Core

  • setup: configure StackGuide context for project type
  • context: show currently loaded context
  • rules: list/search/get/select rule sets
  • knowledge: list/search/get knowledge files
  • review: analyze files/URLs/project code quality

Utility

  • cursor: browse/import community rules
  • docs: fetch/search/list documentation
  • config: save/load/export/import configurations
  • custom_rule: create/update/delete/list project custom rules
  • help: usage help by topic

Advanced

  • generate: boilerplate generation (component/hook/service/test/api/model/util)
  • health: project health score and recommendations
  • analyze: project intelligence (structure/config/dependency/generate/apply)

Workflow

  • workflow: lazy-load raw workflow assets (agents/skills/hooks/commands)
  • init: scaffold .stackguide with stack-aware defaults
  • agent: active workflow executor (status, intake, create_ticket, plan, verify, release)

Active Workflow Details

agent action:"intake"

  • reads ticket from configured tracker
  • returns normalized brief + gaps
  • proposes branch name convention
  • can optionally create a Jira ticket first (createFromDescription:true)

agent action:"create_ticket"

  • creates Jira issues using strict MAIN DESCRIPTION format
  • uses project defaults from .stackguide/config.json (projectKey, issueType)
  • derives summary from first line of MAIN DESCRIPTION when omitted

agent action:"plan"

  • inspects conventions + project shape
  • creates vertical-slice plan
  • returns exactly 3 target tests in structured format

agent action:"verify"

  • executes configured tests/lint/build per layer
  • checks branch naming + commit convention + TDD test budget
  • returns blocker list and final pass/fail report

agent action:"release"

  • checks CI status
  • parses commits since last tag
  • suggests semver impact (major / minor / patch)
  • can create tag and PR/MR (when enabled)

Local Data And Security

init generates:

  • .stackguide/config.json (project workflow config)

Runtime/local artifacts:

  • .stackguide/analysis-cache.json
  • .stackguide/health-history.json

These are machine-local and must not be committed.

Repository Guard

pnpm lint includes:

bash
pnpm run guard:repo

The guard fails if .stackguide artifacts are tracked.

Integrity Hardening

Cache/history persistence includes:

  • boundary/symlink checks
  • file size and entry limits
  • atomic write strategy
  • checksum validation
  • optional HMAC signature using:
bash
STACKGUIDE_INTEGRITY_KEY=<long-random-secret>

For teams, configure STACKGUIDE_INTEGRITY_KEY in local/dev and CI environments for stronger tamper resistance.

Supported Stacks

python-django, python-fastapi, python-flask, react-node, react-typescript, vue-node, nextjs, express, nestjs, laravel, rails, golang, rust, custom.

Development

Use pnpm for all commands.

bash
pnpm installpnpm lintpnpm testpnpm buildpnpm dev

Release And Version Policy

  • Current secure baseline: 4.1.2+
  • Versions from 3.0.0 up to 4.1.1 are deprecated in npm due to local artifact hygiene issues.
  • Use the latest 4.x release in all environments.

Registry Identifiers

  • npm package: @stackguide/mcp-server
  • MCP Registry server: io.github.isucorp-taimiralain/stackguide-mcp

License

GPL-3.0

Source: README.md at commit bd42662

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v4.1.3LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts