Repowise / code-review-graph
code-review-graph alternative: Repowise vs code-review-graph
Choose between a review-oriented graph and wiki, or a broader intelligence layer used across review, planning, onboarding, and maintenance.
code-review-graph produced the most precise retrieval set in the Repowise benchmark, with lower coverage. It is a credible focused option when compact review context matters more than the wider evidence surface.
01 / The short answer
Repowise or code-review-graph?
- code-review-graph
- A local graph and generated wiki designed to provide codebase context for pull-request review agents.
- Repowise
- A persistent codebase-intelligence layer that combines structure, git history, decisions, code health, generated documentation, and MCP retrieval for both people and coding agents.
Pick Repowise when
- You need broader retrieval coverage with file, history, health, and decision evidence.
- You want the index to serve human browsing and workflows beyond pull-request review.
Pick code-review-graph when
- You want a focused local graph and wiki for review-oriented agents.
- High retrieval precision matters more than finding a larger share of relevant files.
02 / Side by side
The facts, row by row.
From each product's own pages, checked August 27, 2026. Sources are listed below.
Scroll sideways to read both products and the notes.
| Question | Repowise | code-review-graph | Note |
|---|---|---|---|
| What is the primary job? | A persistent codebase-intelligence layer that combines structure, git history, decisions, code health, generated documentation, and MCP retrieval for both people and coding agents. | A local graph and generated wiki designed to provide codebase context for pull-request review agents. | Different primary jobs, so the two can run side by side. |
| How does work begin? | Index a repository once; the index then stays current with the code. | code-review-graph focuses its graph and wiki on review. Repowise exposes additional history, health, decisions, risk, and general retrieval workflows. | Pick the starting point that matches how your team works. |
| What can a person inspect? | Documentation, architecture, dependencies, git history, decisions, code health and change risk, in one browsable index. | code-review-graph's own surfaces, as its documentation describes them. | Every Repowise view is open on public repositories without an account. |
| What can an agent use? | MCP server with 10 tools (answers, search, context, symbols, rationale, risk, health), usable from Claude Code, Cursor, Claude.ai and other MCP clients. | code-review-graph's own integrations, as its documentation describes them. | Repowise is an MCP server your agent calls before it edits. |
| Where can it run? | Open source under AGPL-3.0 and self-hostable, or hosted at repowise.dev. | Both are open-source local systems. Setup constraints and agent tool adoption are documented in the benchmark, not hidden. | Plans and data terms from each vendor's own pages. |
03 / What Repowise adds
Every pull request, checked against the whole repo.
The Repowise PR bot reads each change against the dependency graph and git history: what it can break, the hotspots it touches and who owns them. Its risk, coverage and health checks use no model and are free on public repositories. AI review, from credits on Pro and Teams, adds a model pass at about 15 cents a reviewed pull request.
04 / Overlap and difference
Where the workflows meet, and where they split.
Where they overlap
- Both build a code graph and generated repository knowledge for agent consumption.
- Both can support context-aware code review.
- Starting point
- code-review-graph focuses its graph and wiki on review. Repowise exposes additional history, health, decisions, risk, and general retrieval workflows.
- Operating model
- Both are open-source local systems. Setup constraints and agent tool adoption are documented in the benchmark, not hidden.
- Measured
- The benchmark records high precision and lower coverage, plus different adoption across Codex and Claude. Those observations are harness-specific, not a verdict on universal agent behavior.
05 / Check it yourself
Open the product, not a scorecard.
Every link below opens live Repowise output on a public repository. Measured results come straight from the benchmark registry, with their sample attached.
06 / Sources
Checked August 27, 2026.
- code-review-graphPrimary open-source repository for graph, wiki, MCP tools, and setup.Inspect code-review-graph source
- Repowise BenchThe benchmark records high precision and lower coverage, plus different adoption across Codex and Claude. Those observations are harness-specific, not a verdict on universal agent behavior.Inspect benchmark methods and raw results
Questions
What people ask before they try it.
Is Repowise a direct replacement for code-review-graph?
Repowise does a different job. code-review-graph is built for open-source graph and wiki for code review. Repowise is one index of your repository (structure, git history, code health, decisions and documentation), browsed by people and queried by coding agents over MCP. The two can run side by side.
When should a team choose code-review-graph?
You want a focused local graph and wiki for review-oriented agents. High retrieval precision matters more than finding a larger share of relevant files.
When should a team choose Repowise?
You need broader retrieval coverage with file, history, health, and decision evidence. You want the index to serve human browsing and workflows beyond pull-request review.
How does Repowise compare with code-review-graph in benchmarks?
The benchmark records high precision and lower coverage, plus different adoption across Codex and Claude. Those observations are harness-specific, not a verdict on universal agent behavior. Methods, corpora and raw results are public in the Repowise Bench repository.
Put it on your next pull request.
Install the PR bot free on a public repository, or index a repository and read what Repowise finds before you decide.