$ cd ~/projects CodeTend Repository health scans with evidence and finding history ──────────────────────────────────────────────────── A self-hosted platform for continuous, language-agnostic repository health scans, deterministic scores, and tracked technical-debt findings. tags: TypeScript, AI, Web Development, PostgreSQL ──────────────────────────────────────────────────── ## Scan a repository and track what changes Register a repository and run a scan on demand or on a schedule. Specialized agents investigate architecture, duplication, reliability, tests, type safety, dependency health, and security. Their findings include evidence and recommendations. Fixed rules calculate scores and grades from those findings. Finding history distinguishes new, active, improved, resolved, and regressed issues. Mark a result as fixed, a false positive, or an accepted risk with an explanation. A later scan verifies reported fixes. Repository and scanner pages show trends, scan history, and token and cost usage. CodeTend on GitHub https://github.com/michidk/CodeTend ──────────────────────────────────────────────────── ## Take findings into your review process Copy an aggregated fix prompt into your coding agent, or export scan results as Markdown, JSON artifacts, and SARIF. Dependency checks use lockfile versions and vulnerability data; security findings carry evidence for an attack path. For an individual finding, CodeTend can propose a patch in a disposable clone and store it for review. Accept or reject the proposal, then download an approved diff for your normal development process. Generated patches do not write back to the source repository. Optional MCP tools expose repository, scan, and finding workflows to other agents. Architecture and finding lifecycle https://github.com/michidk/CodeTend/blob/main/docs/architecture.md ──────────────────────────────────────────────────── ## Self-hosted proof of concept Docker Compose runs the app, its agent runtime, and PostgreSQL. Configure repository access and an OpenAI-compatible model endpoint, and place a login gate or private network in front of the application. Model calls use that external provider; this is not a fully offline scanner. CodeTend is a proof of concept used on real repositories. Scans have investigation budgets and sample large repositories, so a score does not establish complete coverage. Settings expose budget and concurrency limits, and the operations guide describes deployment and backup. Operations guide https://github.com/michidk/CodeTend/blob/main/docs/OPERATIONS.md Configuration reference https://github.com/michidk/CodeTend/blob/main/docs/configuration.md ──────────────────────────────────────────────────── $ ls ~/pages > about curl -L lohr.dev > projects curl -L lohr.dev/projects > blog curl -L lohr.dev/blog > stats curl -L lohr.dev/stats > contact curl -L lohr.dev/contact ──────────────────────────────────────────────────── https://lohr.dev