Self-hosted AI platform for GitLab

Managing code documentation
shouldn't be a full-time job.

GitKnecht automates code reviews, security scanning, issue triage, and wiki generation — all running on your own infrastructure.

review.py
Complete
1 # GitKnecht reviewed your MR
2 gitknecht.code_review("feature/auth")
3
4 ✓ Quality: 8.2/10 2 vulnerabilities addressed
5 → Comment posted to MR
6 → Wiki regenerated
7 → Issues triaged
100%
Self-hosted & private
6
AI agents working
0
Data leaves your server
1
Command to deploy

How it all connects

GitKnecht integrates with your GitLab instance to automate the entire development workflow — from code review to documentation.

The Problem

Every day, your team creates merge requests, files issues, and updates documentation. But code reviews get delayed, security vulnerabilities slip through, issues pile up unsorted, and your wiki becomes outdated the moment you push code. It's a full-time job just keeping everything in sync.

The Solution

GitKnecht runs alongside your GitLab instance as a self-hosted service. It connects via the GitLab API, monitors your projects for new merge requests and issues, and deploys AI agents that work in the background — reviewing code, scanning for security issues, classifying and deduplicating issues, and keeping your wiki documentation automatically updated.

How It Works

When you push code and create a merge request in GitLab, GitKnecht's polling service detects the change within seconds. It clones your project, analyzes the diff with an LLM-powered review engine, and posts structured feedback directly as an MR comment. Simultaneously, it scans open issues for duplicates using semantic tag matching, regenerates wiki pages that match changed files, and answers team questions grounded in your actual documentation.

GitLab Integration

Connects via GitLab API to monitor MRs, issues, and wiki changes in real-time.

Local State

Maintains a local database of projects, reviews, issues, and wiki pages for fast access.

AI Agents

Multiple specialized agents handle code review, security scanning, issue triage, and wiki generation.

Private & Secure

All data stays on your infrastructure. API keys encrypted at rest. Zero data leaves your server.

Architecture Overview

GitLab Instance

Source of truth — projects, MRs, issues, and wiki pages

Polling Service

Monitors GitLab for new MRs and issues on configurable intervals

Review Engine

LLM-powered code analysis with multi-model fallback and structured output

Triage Agent

Classifies issues, extracts semantic tags, deduplicates using fuzzy matching

Wiki Engine

Diff-aware regeneration keeps documentation synced with code changes

Chat Interface

Documentation-grounded Q&A — ask questions, get answers from your wiki

Everything your team needs

Six AI agents working in concert to keep your codebase clean, secure, and well-documented.

AI Code Reviews

Automated merge request reviews powered by LLMs. Detects bugs, anti-patterns, and quality issues with structured feedback — posted directly as MR comments.

Security Scanning

Identifies vulnerabilities, hardcoded secrets, and dangerous anti-patterns in every diff. Your code stays secure without manual effort.

Issue Triage

LLM-powered issue classification in any language. Automatic deduplication using semantic tag matching. Bugs and features sorted intelligently.

Auto Wiki Generation

When code changes, your wiki updates automatically. Diff-aware regeneration keeps documentation in sync with your codebase.

Documentation Chat

Ask questions about your project and get answers grounded in your actual wiki documentation. Contextual, accurate responses.

Smart Polling

Configurable polling supervisor with MR-level blocking prevents duplicate reviews. Multiprocessing pool scales to your needs.

MR Conflict Detection

Scans for potentially conflicting merge requests based on branch keywords and overlapping components. Know before you merge.

Multi-Model Fallback

Supports multiple LLM providers (OpenAI, Anthropic, OpenRouter, Azure). Automatic fallback keeps reviews flowing.

Encrypted Secrets

All API keys and credentials are encrypted at rest. Your infrastructure secrets never leave your server in plaintext.

From push to production

GitKnecht works silently in the background, automating the tedious parts of development.

01

Push your code

Create a merge request in GitLab as usual. GitKnecht detects it via configurable polling or webhooks.

02

AI reviews your code

The review engine analyzes your diff with LLM-powered analysis, checking for quality, security, and best practices.

03

Issues get triaged

Open issues are classified, tagged, and deduplicated. No more duplicate bug reports or unsorted feature requests.

04

Wiki stays updated

Changed files trigger wiki regeneration. Your documentation stays in sync with your codebase automatically.

05

Merge with confidence

Review the AI feedback, address any findings, and merge knowing your code has been thoroughly vetted by AI — all without leaving GitLab.

Built with
Python
FastAPI
PostgreSQL
SQLAlchemy
Alembic
Next.js
TypeScript
Docker
RegCode AI
LiteLLM
Pytest
Ruff
GitLab API
OIDC Auth
Python
FastAPI
PostgreSQL
SQLAlchemy
Alembic
Next.js
TypeScript
Docker
RegCode AI
LiteLLM
Pytest
Ruff
GitLab API
OIDC Auth

Ready to get started?

Self-hosted. AI-powered. Zero data leaves your server.

Launch GitKnecht