The AI-debt radar alternative.
GitClear measures AI's aggregate effect on code quality across the industry. repowise tells you which of your code an AI agent wrote, whether it is healthy, and who owns it, all open source and self-hostable.
GitClear proved that AI assistants change how code gets written: more copy-paste, less refactoring, more churn, more duplicated blocks. The harder question is what that means for your repositories, this quarter, file by file.
GitClear answers the industry-trend question with research across hundreds of millions of lines. repowise answers the operational one: which AI-written code in your repo is also a low-health hotspot owned by a single person. It fuses per-agent provenance with a defect-validated health score and bus-factor ownership, the AI-debt radar no one else ships end to end, computed from git history alone.
Which one is right for you?
Choose repowise if
- You want to know which of your own code AI wrote, not an industry average
- You want that provenance tied to a defect-validated health score and bus-factor ownership
- You want risk management for AI-era code, not per-developer productivity surveillance
- You want code health, an auto-generated wiki, decisions, and agent-native MCP in one open tool
- You want open source and self-hostable, with every heuristic inspectable
Choose GitClear if
- You want broad engineering-productivity analytics across 65+ metrics (Diff Delta)
- You want established SEI dashboards and per-contributor team stats
- You want the credibility of a viral, large-scale industry research dataset (150M+ lines)
- You want churn, rework, and copy-paste diff analytics as a standalone product
repowise vs GitClear
| Capability | repowise | GitClear |
|---|---|---|
| Per-agent provenance of your own codeGitClear measures AI's aggregate industry effect, not per-agent attribution of your commits | Included | Not included |
| Provenance fused with health and ownership (AI-debt radar) | Included | Not included |
| Defect-validated code-health scorerepowise: ROC AUC 0.74, reproducible on your repo | Included | Not included |
| Three co-equal health signals (defect, maintainability, performance)GitClear has no comparable code-health scoring on any pillar | Included | Not included |
| No account or API key for the first scanrepowise: pip install repowise; GitClear requires connecting your git host first | Included | Not included |
| Churn, rework, and copy-paste diff analyticsGitClear's Diff Delta and code-quality stats are deeper here | Partial support | Included |
| Broad engineering-productivity metrics (65+) | Not included | Included |
| Large-scale industry AI-code research datasetGitClear's 150M+ line studies are a real credibility asset | Not included | Included |
| Hotspots, ownership, and bus factor | Included | Partial support |
| Open source and self-hostable | Included | Not included |
| Auto-generated wiki and documentation | Included | Not included |
| Architectural decision records | Included | Not included |
| Agent-native MCP context (overview, answers, risk, why) | Included | Not included |
| Risk framing, not per-developer surveillanceGitClear includes per-contributor productivity stats | Included | Partial support |
Self-assessed against publicly documented features as of June 2026. A dash means partial or limited support. Vendor capabilities change, so please verify against GitClear's current docs before deciding.
From industry trend to your repo, today.
The same AI-code concern GitClear popularized, made operational on your codebase and fused with the signals that turn it into risk you can act on.
Which AI-written code is a single-owner hotspot
repowise fuses three signals no other tool combines on one surface: per-agent provenance of your code, a defect-validated health score, and bus-factor ownership. The result is a directional risk read, not a productivity ledger: where AI-written code is also low-health and concentrated with one person.
- Provenance derived from git history, no IDE plugins, no per-developer surveillance
- Tied to a 1 to 10 health score from 21 deterministic markers
- Bus factor: files owned more than 80% by a single author
- Reads commits, not people: codebase-level risk, not engineer ranking
A health score proven to find your bugs
GitClear measures change quality with Diff Delta. repowise adds the missing layer: a code-health score validated against real defect labels and reproducible on your own repo, so you can confirm it finds your bugs rather than taking a vendor's word for it.
- Cross-project ROC AUC 0.74, up to 0.90 per repo
- 2.3x more defects under a fixed review budget vs a leading tool
- On a typical repo, 16 of the 20 worst files had a recent bug fix, 3.3x the baseline
- Zero LLM in scoring: under 30 seconds on a 3,000-file repo
Provenance, health, docs, decisions, and agent context
GitClear is a closed-source analytics SaaS. repowise puts AI provenance and health alongside an auto-generated wiki, architectural decision archaeology, git intelligence, and ten MCP tools, all open source and self-hostable, so the same index serves your AI-risk goals and your AI agents.
- Auto-generated wiki, rebuilt on every commit
- Architectural decisions mined from eight sources
- MCP tools for Claude Code, Cursor, Cline, and Codex
- AGPL-3.0: inspect, fork, self-host, zero telemetry
Git intelligence is one input, not the whole score.
GitClear analyzes commits and diffs. repowise reads the same git history but folds it into a validated code-health model with no separate account or LLM key required to start.
Defect risk
The primary 1-to-10 score, from 21 markers validated against real defects (ROC AUC 0.74), git signals like hotspots and ownership included.
Maintainability
8 markers for readability and change-cost smells, reported as its own co-equal score rather than folded into the defect number.
Performance
20 markers for static I/O-in-loop and N+1 risk shapes, a third co-equal pillar GitClear's git-only lens does not cover.
None of this needs a connected git-host account to start. pip install repowise, then repowise init --yes --no-prose builds the dependency graph, git history, agent provenance, and all three health pillars locally, with zero LLM calls.
The honest version
GitClear is a strong, well-built product, and there are places it leads. Its Diff Delta metric and 65+ engineering metrics make it far broader on the team-analytics and developer-productivity axis than repowise aims to be, with mature churn, rework, and copy-paste diff analytics and per-contributor stats. Its viral, large-scale industry research, analyzing hundreds of millions of lines to show how AI assistants drive copy-paste up and refactoring down, is a genuine credibility asset that repowise does not have. And its established software-engineering-intelligence dashboards serve a team-management need repowise does not target. If broad SEI analytics and an industry research dataset are your priority, GitClear is a strong choice. repowise wins when you want to know about your own code, fuse AI provenance with defect-validated health and ownership, and keep it open, self-hostable, and free of per-developer surveillance.
What repowise costs.
The repowise core is open source under AGPL-3.0 and free to self-host, with every heuristic public. Hosted tiers: Free for public repos, Pro at $15 per month, Teams at $60 per month, and custom enterprise licensing when you need it. GitClear's pricing changes, so verify it on their site, then compare it with a tier you can read line by line.
Questions, answered
Is repowise a good GitClear alternative?
It depends on what you are trying to answer. GitClear is a software engineering intelligence platform built on its Diff Delta metric, with 65+ engineering metrics and viral, large-scale research on how AI assistants change code quality across the industry. repowise answers a different and more specific question: of your own code, how much did an AI agent write, is that code a low-health hotspot, and who owns it. If you want that AI-debt radar fused with a defect-validated health score and bus-factor ownership, open source and self-hostable, repowise is the better fit. If you want broad per-contributor productivity analytics and an industry research dataset, GitClear is stronger on that axis.
Does repowise show how much of our code AI wrote?
Yes. repowise derives agent provenance from your git history alone, with no IDE plugins and no per-developer instrumentation, and ties it to the code-health score and to ownership. So you see not just that AI wrote a chunk of a file, but whether that file is a low-health hotspot and whether a single person owns it. Industry-wide, roughly 41 to 42% of code is AI-written in 2026, but repowise reports the share in your repositories, not an industry average.
Is repowise developer-surveillance?
No. repowise reads commits, not people. Agent provenance is a directional risk signal at the codebase level, not a precise per-developer productivity ledger. The question it answers is which AI-written code is also a low-health hotspot owned by a single person, which is risk management for AI-era codebases, not a productivity score for ranking engineers. We deliberately avoid the productivity-surveillance framing because teams and individual engineers increasingly distrust it.
How is repowise different from GitClear's Diff Delta?
Diff Delta is GitClear's measure of durable change per commit, filtering out noise, churn, and copy-paste to estimate meaningful work and developer productivity. It is a broad team-analytics metric. repowise is not a productivity measure. It pairs per-agent provenance of your code with a defect-validated code-health score (1 to 10 per file from 21 deterministic markers) and bus-factor ownership, so the output is where the risk is concentrated, not how productive a contributor was.
Is repowise open source and self-hostable?
Yes. The repowise core is open source under AGPL-3.0, so every marker, weight, and scoring rule is public and inspectable, and you can self-host the whole platform with zero telemetry and code that never leaves your infrastructure. GitClear is a closed-source SaaS.
Does repowise have a code-health score?
Yes, and it is the part GitClear does not have. repowise scores every file 1 to 10 from 21 deterministic markers (complexity, nesting, cohesion, clones, change entropy, co-change scatter, ownership dispersion, prior-defect history, and more), with no LLM, in under 30 seconds on a 3,000-file repo. It is defect-validated: cross-project ROC AUC 0.74 (95% CI 0.68 to 0.79, up to 0.90 per repo), and 2.3x more defects surfaced under a fixed review budget than a leading commercial tool on the same 2,770 files.
Is repowise's health score one number or several?
Three co-equal signals: defect risk (the 1-to-10 score above), maintainability (8 markers for readability and change-cost smells), and performance (20 markers for static I/O-in-loop and N+1 risk shapes). GitClear does not publish a comparable defect-validated score at all, on any pillar; its metrics measure change volume and durability, not code health.
Do I need an API key to start scoring code health?
No. pip install repowise, then repowise init --yes --no-prose builds the dependency graph, git history, agent provenance, and all three health pillars with zero LLM calls. GitClear requires connecting your git host and an account before it reports anything; repowise's deterministic layer needs neither.
How much of your code AI wrote, and whether it is healthy.
Hotspots, ownership, coupling, bus factor.
How hotspots, ownership, and coupling are computed from git history alone.
Where repowise's git intelligence ranks next to GitClear.
Finding the risky files churn and complexity both agree on.
Files that change together but share no import.
Measuring single-person risk from git history alone.
The defect-validated score, in depth.
Health, ownership, and AI-debt for leaders.