Everything you need to know what breaks

Graph-native analysis

Not grep. Not embeddings. An actual call graph.

Arbor uses Tree-sitter to parse your source into ASTs, then builds a directed dependency graph via petgraph. Every edge is a real call — function calls, imports, class extensions, interface implementations.

When you change a function, Arbor does a BFS from that node and finds every transitive caller. No fuzzy matching. No hallucinations. If A → B → C and you change C, Arbor reports A and B. Always.

Full — ownership + lifetime graph

Full — imports, interfaces, generics

Full — ESM + CJS module graph

Full — imports, class hierarchy, decorators

Full — package-level dependency graph

Full — class hierarchy, interface chains

Full — include graph + function calls

Full — templates, namespaces

Full — namespaces, LINQ expression tracking

Partial — class hierarchy, top-level fns

Partial — protocols, extensions

Partial — class + module hierarchy

Partial — namespace + class graph

Partial — Flutter widget tree

Sub-30s analysis

Webhooks trigger the analysis pipeline immediately. Most results post to the PR within 20–30 seconds.

GitHub App

Install in 30 seconds. No CI changes. No config files. Just install and open a PR.

Commit status gates

Set HIGH blast radius as a merge-blocking status check on your branch protection rules.

Dashboard & history

Every analysis is saved. Track risk trends per repo, per branch, per engineer.

No code sent to AI

Pure static analysis. Your code never leaves your infrastructure for AI processing.

Slack notifications

Get a Slack message with the blast radius summary when a HIGH-risk PR is opened.