MCP Server Usage Guide

This project has MCP servers configured in the workspace

.mcp.json

Available Servers

Git & GitHub Tool Selection Guide

Three tools available for version control and GitHub operations. Choose based on the task:

git push

gh pr

gh issue

gh issue

gh issue

gh search

gh api

git

⚠️ IMPORTANT: The GitHub MCP server frequently has authentication failures ("Bad credentials" errors) even when tokens are configured. Always prefer the

gh

CLI via Bash for GitHub operations (issues, PRs, searches). The

gh

CLI is pre-authenticated and reliable.

Decision Flowchart

Is it a GitHub.com operation (issues, PRs, remote files)?
├─ Yes → Use Bash `gh` CLI (gh issue, gh pr, gh api)
│        └─ GitHub MCP is unreliable due to auth issues
└─ No → Is it a local git operation?
        ├─ Yes → Use Git MCP
        │        └─ Unless it's push/pull/fetch → Use `git` in Bash
        └─ No → Use Bash

Git MCP Server

Use for local repository operations:

mcp__git__git_status      # Working tree status
mcp__git__git_diff        # View changes (staged/unstaged)
mcp__git__git_log         # Commit history
mcp__git__git_commit      # Create commits
mcp__git__git_branch      # List branches
mcp__git__git_checkout    # Switch branches
mcp__git__git_add         # Stage files

Limitations: Cannot push, pull, fetch, or perform remote operations.

GitHub MCP Server

⚠️ NOT RECOMMENDED: The GitHub MCP server frequently fails with "Bad credentials" authentication errors. Use the

gh

CLI via Bash instead for all GitHub operations.

Available tools (but prefer

gh

mcp__github__create_issue           # Use: gh issue create
mcp__github__update_issue           # Use: gh issue edit
mcp__github__add_issue_comment      # Use: gh issue comment
mcp__github__list_issues            # Use: gh issue list
mcp__github__create_pull_request    # Use: gh pr create
mcp__github__get_pull_request       # Use: gh pr view
mcp__github__search_code            # Use: gh search code
mcp__github__get_file_contents      # Use: gh api repos/OWNER/REPO/contents/PATH

Limitations: Unreliable authentication, no milestone CRUD.

Bash git/gh CLI (Preferred for GitHub)

The

gh

For detailed issue management (epics, sub-issues, milestones, GraphQL API), see the

/github-issues

skill.

# Issues (preferred over GitHub MCP)
gh issue list --search "keyword"           # Search issues
gh issue list --state all --limit 30       # List all issues
gh issue view 123                          # View issue details
gh issue create --title "Title" --body "Body"
gh issue edit 123 --add-label "bug"
gh issue comment 123 --body "Comment text"

# Pull Requests (preferred over GitHub MCP)
gh pr list                                 # List open PRs
gh pr view 123                             # View PR details
gh pr create --title "Title" --body "Body"
gh pr merge 123 --squash

# Search (preferred over GitHub MCP)
gh search issues "keyword repo:owner/repo"
gh search code "function_name repo:owner/repo"

# API access for anything else
gh api repos/OWNER/REPO/milestones
gh api repos/OWNER/REPO/contents/path/to/file

# Remote git operations (not in Git MCP)
git push origin branch-name
git pull origin main
git fetch --all

# Complex git operations
git rebase -i HEAD~3
git cherry-pick abc123
git stash push -m "message"

Context7 Usage

Use Context7 to fetch up-to-date documentation for libraries and frameworks:

# Get documentation for a library
resolve-library-id: "react"
get-library-docs: { libraryId: "/react/react", topic: "hooks" }

Best for:

• Checking current API signatures
• Understanding library-specific patterns
• Verifying framework best practices

Knip Usage

Use Knip to detect dead code and unused dependencies:

# Run knip analysis
run_knip: { cwd: "/path/to/project" }

# Get knip documentation
get_knip_docs: { topic: "configuration" }

Best for:

• Finding unused exports before refactoring
• Identifying dead code after feature removal
• Cleaning up unused dependencies

Memory Server Usage

The memory server maintains project knowledge across sessions. Use it to:

Query existing knowledge:

mcp__memory__search_nodes("ruleset")     # Find ruleset architecture info
mcp__memory__search_nodes("technical")   # Find known technical debt
mcp__memory__open_nodes(["KeyFiles"])    # Get key file locations

Store new knowledge:

mcp__memory__create_entities([...])      # Add new architectural concepts
mcp__memory__add_observations([...])     # Add details to existing entities
mcp__memory__create_relations([...])     # Link concepts together

When to update memory:

• After making significant architectural decisions
• When discovering important patterns or gotchas
• After resolving tricky bugs (document the solution)
• When adding new major features

Sequential Thinking Usage

Use

mcp__sequentialthinking__sequentialthinking

• Debugging complex ruleset merge issues
• Planning multi-step refactors
• Working through character creation edge cases
• Designing new edition support
• Any problem requiring step-by-step reasoning with revision