CAT
/MCP
SkillsMCPMarketplacesDigestToolsAdvertise

This week in Claude

Every Monday: Claude Code, Agent SDK, MCP, and the Anthropic platform moves worth your time.

Skills by Category
Frontend DevelopmentBackend & APIsTesting & QASecurityDevOps & CI/CDGit & Pull RequestsDocumentationCode Review & QualityAI & Agent BuildingSkill Development
MCP Servers by Category
Sales & MarketingWeb & Browser AutomationDatabasesAI & LLM ToolsCloud & InfrastructureCommunication & MessagingDeveloper ToolsDesign & CreativeDocuments & KnowledgeSearch & Web Crawling
Marketplaces by Category
AI Agents & OrchestrationLLM IntegrationDevelopment ToolsFrontend & UIBackend & APIsDatabasesTesting & Code QualityDevOps & CloudSecurity & ComplianceGit & Version Control

Cross AI Tools

Discover Claude Code plugins, extensions, and tools. Automatically updated directory of Anthropic Claude AI marketplaces with development tools, productivity plugins, and integrations.

Resources

  • Browse Skills
  • Browse MCP Servers
  • Browse Marketplaces
  • Plugins Reference

Community

  • About
  • Tools
  • Feedback
  • Privacy Policy
  • Advertise

Built for the Claude Code community with Claude Code by @mertduzgun

Independent project, not affiliated with Anthropic

Git Courer

alejandro-m-p/git-courer
32registry active
Summary

A full-featured Git interface built specifically for LLM agents, exposing 22 operations from commits and rebases to PR reviews and automated releases. Instead of dumping raw diffs into your LLM, it parses ASTs with go-tree-sitter, classifies change types deterministically in Go, then hands structured context to the model for message generation. Every destructive operation gets an automatic backup. The pr-review tool runs tests, detects conflicts with AST annotations, and checks divergence in one shot. Ships with a CLI release command that groups commits by configured areas and calculates semver bumps locally. Auto-configures 14 clients including Claude, Cursor, and Zed. Reach for this when you want Git operations that understand code structure, not just text diffs.

CodeRabbit
CodeRabbit
AI writes the code. CodeRabbit catches the slop.
Try For Free →
Keep your Mac awake
Keep your Mac awake
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →
Context.devContext.dev
Context.dev
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
Make your agent a DeFi expert
Make your agent a DeFi expert
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Make money from your Skills
Make money from your Skills
On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →
AppSignal
AppSignal
Monitor with ease. Code with confidence.
Start Free Trial →
CodeRabbit
CodeRabbit
AI writes the code. CodeRabbit catches the slop.
Try For Free →
Keep your Mac awake
Keep your Mac awake
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →
Context.devContext.dev
Context.dev
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
Make your agent a DeFi expert
Make your agent a DeFi expert
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Make money from your Skills
Make money from your Skills
On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →
AppSignal
AppSignal
Monitor with ease. Code with confidence.
Start Free Trial →
Gemini_Generated_Image_g9lcw7g9lc

Release Build MIT License

Issues & Bugs: @blak0p/git-courer/issues · Discussions: @blak0p/git-courer/discussions


Quick Links

DocDescription
WebVisit the official website
RoadmapWhat's coming next and the strategic vision
ArchitectureCodebase structure, patterns, and how to add features
TroubleshootingFix: Ollama not running, MCP not detected, permission errors
MCP ClientsAll 14 supported clients, config formats, manual setup
Config OptionsAll ~/.config/git-courer/config.yaml and .git/git-courer/config.json settings
CommandsComplete reference for all 22 MCP tools
Models GuideTested models, token usage, and which one to pick
ContributingSetup, running tests, and how to collaborate

Install

go install github.com/blak0p/git-courer@latest
curl -fsSL https://raw.githubusercontent.com/blak0p/git-courer/main/scripts/install.sh | sh

That's it. It installs the binary and auto-configures every AI tool it detects on your machine.

Requirements: Git · llm local

Homebrew (macOS / Linux):

brew install blak0p/tap/git-courer

Windows

# Windows (PowerShell)
irm https://github.com/blak0p/git-courer/releases/latest/download/git-courer_windows_amd64.tar.gz | tar -xz -o git-courer.exe
.\git-courer.exe setup

Manual install:

# macOS / Linux
curl -fsSL https://github.com/blak0p/git-courer/releases/latest/download/git-courer_$(uname -s | tr '[:upper:]' '[:lower:]')_$(uname -m).tar.gz | tar -xz -C /usr/local/bin git-courer
chmod +x /usr/local/bin/git-courer
git-courer setup

git-courer

The only MCP git server that understands your code before it commits.

Most tools wrap git diff | llm "write a commit". git-courer does something different: it parses your AST, builds a dependency graph, classifies the change type in Go without touching the LLM, an[...]

One staged area = one commit. Always.


Why it's different

1. Commit type determined in Go, not by the LLM

git-courer uses go-enry and go-tree-sitter to parse your AST and apply deterministic rules — new public function, modified signature, deleted symbol, breaking change. The LLM writes the message,[...]

2. Dependency graph

Before committing, git-courer builds a graph of what your staged files affect across the codebase. Real scope, not just which files were touched.

3. Local + cloud working together

Go and the local LLM handle everything they can. Your cloud LLM receives structured commit context — WHY + WHAT — instead of raw code.

4. Commits as LLM context

A git-courer commit is a compressed, structured summary any LLM can consume directly. Better input, fewer tokens, fewer hallucinations.

5. Automatic backup on every mutation

Every write operation creates a backup before executing. One command undoes anything.

6. The complete git pack — one server, nothing else needed

Why install one tool for commits, another for releases, another for PR checks? git-courer covers the full cycle:

Read: status, diff, history, blame, pr-review
Write: commit, amend, revert, stage, reset, stash
Branch: branch, merge, rebase, cherry-pick, tag
Utility: sync, backup, undo, remotes, config

All returning structured JSON. No pager hangs, no text parsing, no extra tools.


Workflows

Commit

Preview the change → review proposed commits → apply. git-courer splits your staged files into atomic commits by dependency graph automatically.

Before any PR or merge

Call pr-review — a pre-PR gate that runs in one shot:

  1. Tests — runs test_command from .git/git-courer/config.json (e.g. go test ./...)
  2. Conflicts — detects merge conflicts with the target branch and returns AST-annotated hunks ([NEW_FUNC], [MOD_SIG ⚠BREAKING])
  3. Diff stats — files changed, additions, deletions
  4. Divergence — ahead/behind count, mergeable status

If it's not green, you don't merge. Set test_command via git-config SET_TEST_COMMAND or edit .git/git-courer/config.json directly.

Release (CLI)

Run git-courer release. The CLI reads commits from .git/git-courer/branches/ (captured during each git-commit APPLY on this branch) and also from refs/courer/* — per-branch commit blobs that survive squash merges and deleted branches

Undo

Every destructive operation has an automatic backup. One command restores the previous state.


How it works

You: "commit my changes"
        ↓
AI delegates to git-courer (via MCP)
        ↓
Go reads AST + dependency graph → classifies type deterministically
        ↓
Local LLM writes the human-readable message from the annotated diff
        ↓
Security scan (5 layers) → auto-backup → commit
        ↓
"✓ feat(auth): add OAuth2 token refresh"

For the full list of tools: docs/commands.md
For workflow details: docs/workflows.md


Supported Tools

ToolAuto-configured
Claude Code✓
Cursor✓
Windsurf✓
OpenCode✓
Cline✓
Roo Code✓
VS Code✓
Claude Desktop✓ macOS/Win only
Continue✓
Zed✓
Codex✓
Gemini CLI✓
pi✓
Antigravity✓

Run git-courer mcp setup to configure all detected tools at once, or git-courer mcp setup <client> for a specific one.

Interactive TUI

Run git-courer with no arguments to launch the interactive installer:

Interactive TUI

The TUI walks you through 4 steps:

  1. MCP Configuration — select which AI tools to configure (auto-detects installed clients)
  2. General Settings — configure your LLM backend, model, and project context
  3. Review — preview your config before saving
  4. Finish — config saved to ~/.config/git-courer/config.yaml

You can also update the binary or uninstall directly from the menu. Navigation: j/k or arrow keys, esc to go back.

Commands

git-courer runs as an interactive TUI when launched without arguments. It also provides MCP server and management commands:

CommandDescription
git-courerLaunch interactive TUI (requires terminal)
git-courer mcpRun MCP server
git-courer mcp setupConfigure all detected AI tools
git-courer mcp setup <client>Configure a specific tool (e.g. cursor)
git-courer releaseAutomated semver releases and changelogs (CLI only)
git-courer removeRemove git-courer from the current project
git-courer uninstallUninstall the binary globally
git-courer updateUpdate to the latest version
git-courer versionShow current version

MCP Tools

For the 22 MCP tools and their arguments, see docs/commands.md.

Configuration

git-courer uses two config levels:

Global (~/.config/git-courer/config.yaml) — personal settings: LLM backend, model.

Per-project (.git/git-courer/config.json) — committable, shared with team. Stores description, areas, test_command, excluded. Better results = edit this file per project.

All options: docs/config.md

Background Jobs

When you call git-commit PREVIEW, the server may return immediately (FAST) or start a background job (SLOW). In the SLOW case:

  1. PREVIEW returns {status:"processing", job_id}
  2. The agent must call git-commit STATUS with that job_id to poll
  3. When STATUS returns {status:"done"}, the plan is ready
  4. Then call git-commit APPLY with the same job_id

The agent continues working — it does NOT block waiting for the job.

Every successful commit via git-commit APPLY is captured to .git/git-courer/branches/<branch>/commits.json and a refs/courer/<branch> blob is created. When you later run git-courer release, it reads commits from both the local store and the refs — so squash-merged PRs still produce accurate changelogs

Why this matters: Git history is frequently rewritten — PR squashes, rebases, force-pushes — destroying the real commit narrative. The CommitStore preserves every commit message as it was[...]

Troubleshooting

Having issues? Check docs/troubleshooting.md for:

  • Ollama not running / model not configured
  • MCP not detected by your AI tool
  • Permission errors during install
  • Secrets detected in commits (false positives)

MCP config file locations: docs/mcp-clients.md


FAQ

Who decides the commit type? Go, via AST analysis. The LLM only writes the human-readable message.

Do I need Ollama? You need some LLM backend. Ollama is the recommended default, but git-courer works with any OpenAI-compatible server: LM Studio, vLLM, LocalAI, or a custom endpoint. Without a configured backen[...]

Is my code sent anywhere? No. Everything runs on your machine — git-courer, Ollama, your data.

Who decides the version number in a release? Go, not Ollama. Version is calculated from commit types (feat: → minor, feat!: → major). Ollama only writes the human changelog.

My tool isn't listed. Open an issue: @blak0p/git-courer/issues. If it supports MCP, adding it is usually a few lines.

How do I mark a breaking change? Use ! after the commit type (feat!:) or include BREAKING CHANGE: in the body. git-courer picks this up automatically for version bumping and changelog generation.


Used by

WhoStackModel
@blak0pOpenCode · Linux · RX 9070 XTgemma4:e4b

Running it? Drop your setup in Discussions — I'll add you to this table.


Contributing

Want to collaborate? Here's everything you need:

Architecture & Codebase

Read docs/architecture.md for:

  • Directory structure and tech stack
  • Hexagonal architecture patterns
  • Key packages and their responsibilities
  • How to add a new feature
  • Testing approach

Reporting Bugs

Found a bug? Open an issue: @blak0p/git-courer/issues

Include:

  • Your OS and git-courer version (git-courer version)
  • AI tool you're using (Claude Code, Cursor, etc.)
  • Steps to reproduce
  • Relevant logs or error messages

How to Collaborate

  1. Read the docs: Start with docs/architecture.md and CONTRIBUTING.md
  2. Pick an issue: Check @blak0p/git-courer/issues for good first issue labels
  3. Discuss: Use @blak0p/git-courer/discussions for questions or feature ideas
  4. Submit PR: Follow conventional commits (feat:, fix:, chore:)

Adding a New MCP Client

If your AI tool supports MCP but isn't listed, adding it is usually 5 lines of code in internal/installer/mcp_config.go. See docs/mcp-clients.md for the format.

Featured
CodeRabbit
CodeRabbit
AI writes the code. CodeRabbit catches the slop.
Try For Free →
Keep your Mac awake
Keep your Mac awake
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →
Context.devContext.dev
Context.dev
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
Make your agent a DeFi expert
Make your agent a DeFi expert
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
Make money from your Skills
Make money from your Skills
On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →
AppSignal
AppSignal
Monitor with ease. Code with confidence.
Start Free Trial →
Categories
AI & LLM ToolsDeveloper Tools
Registryactive
UpdatedMay 30, 2026
View on GitHub

Related AI & LLM Tools MCP Servers

View all →
SkillFM LLM Cost Optimizer

io.github.ericm1018/skillfm-llm-cost-optimizer-openai-anthropic-usage

LLM cost optimizer for OpenAI, Anthropic, token usage, BYOK, and SkillFM Beacon audits.
Llm Orchestration Agent

io.github.mikerawsonnz/llm-orchestration-agent

Run a prompt through a LangChain (system + human) chain over Gemini on Vertex AI; optional LangSmith
Authenticated Llm Agent

io.github.mikerawsonnz/authenticated-llm-agent

JWT-gated LLM gateway: authenticate (bcrypt/JWT), then run a LangChain-on-Vertex Gemini completion.
Copilot Memory MCP

labforgedev/copilot-memory-mcp

Persistent semantic memory for AI agents using local ChromaDB vector search. No cloud required.
1
Agent Prompt Injection Firewall Mcp

csoai-org/agent-prompt-injection-firewall-mcp

The WAF for agents. Pattern-based + heuristic firewall scans prompts, RAG documents, tool argume...
Authenticated Multi Llm Agent

io.github.mikerawsonnz/authenticated-multi-llm-agent

Google-OAuth-gated LLM gateway: verify a Google ID token, then run a Gemini (Vertex AI) completion f