Writing code got easy.
Keeping a codebase healthy didn't.
Copilot writes the code. Developers merge the PR. Nobody checks what it did to the repo. Octokraft does.
The problem
AI makes code generation fast.
It doesn't make your codebase better.
The pattern is the same everywhere. AI assistants generate PRs in minutes. Developers merge dozens a week. Each one looks fine individually. But nobody tracks what all those changes are doing to the codebase as a whole.
A developer opens a PR. The diff is 400 lines. Did they write it, edit it, or just hit accept? The reviewer can't tell. The more code ships this way, the less anyone truly reviews.
Half the team writes error handling one way, the other half does something different. The "standard" is whatever the senior devs know. AI assistants don't know it at all.
Each commit looks fine on its own. But over weeks the repo gets harder to change, and nobody can point to the moment it got bad. Quality just drifts.
Code health
Quality degrades gradually. Then all at once.
A codebase doesn't rot overnight. It rots across hundreds of commits, each one fine on its own. By the time someone says "this repo is a mess," the damage is months old. Octokraft scores security, tests, complexity, duplication, dead code, consistency, runtime risks, and compliance on every assessment, so the drift shows up before anyone feels it.
Explore the full health dashboardIssues
Ranked by what breaks production, not what breaks a linter.
Security vulnerabilities surface before style nitpicks. Runtime risks rank higher than naming complaints. Findings link to the exact file and line. Each one includes enough context to decide whether to fix it now or later.
Explore the full issue listARCHITECTURE
The architecture nobody documented is still in the code.
Octokraft reads the module boundaries, the coupling patterns, the conventions that emerged organically. Architecture reviews track whether the system is getting tighter or falling apart. ADRs stay connected to the code they describe.
Explore livePULL REQUESTS
A renamed function breaks callers in three other services.
Octokraft checks each PR against the full codebase. A refactor that looks clean in the diff might introduce scope drift or violate patterns established elsewhere. The verdict comes before the merge button.
Explore liveHow it works
Install the GitHub app. Get a health score.
Install the GitHub App
Pick your repos. Octokraft detects your stack, runs the analyzers, builds a code graph, and gives you a health score.
PRs get real context
A renamed function gets checked against every caller. A new endpoint gets compared to your existing patterns. Not just the changed lines.
Track health over time
You'll know when quality drops, when patterns drift, and when the architecture starts to shift. One dashboard for all your repos.
Integrations
Plugs into the tools already on the machine.
MCP server for Claude Code, Cursor, and Windsurf. Findings push to GitHub Issues, Linear, or Jira. LLM provider is configurable: OpenAI, Anthropic, Azure, or a self-hosted model.
{
"created_at": "2026-03-11T21:01:21+01:00",
... +50 lines (ctrl+o to expand)
... +120 lines (ctrl+o to expand)
... +30 lines (ctrl+o to expand)
Preflight
Need an answer before the PR exists?
Preflight is the local-first branch readiness CLI. It watches the repo on your machine, explains what changed, highlights blast radius and test gaps, and keeps historical review context tied to the exact snapshot that produced it.
Local snapshot identity
Branch, base, fingerprint, and review scope stay attached to the current worktree.
Before-PR triage
Catch blockers, scope drift, and risky fan-out before anybody opens GitHub.
Public releases
Installers, docs, and release bundles ship from a public GitHub repo.
Unix
curl -fsSL https://raw.githubusercontent.com/octokraft/preflight/main/scripts/install.sh | sh
PowerShell
irm https://raw.githubusercontent.com/octokraft/preflight/main/scripts/install.ps1 | iex
The public repo hosts installers, docs, release notes, and checksums. Source development stays private.
Built for real systems
Works across all your repos.
Most tools look at one repo at a time. Octokraft builds a graph across your whole system. Dependencies, symbols, cross-repo relationships. One dashboard that shows how your services actually connect.
Cross-repo health
Compare repos, track trends, find the one that's dragging everything down.
Impact propagation
A change in service A affects consumers in service B. See it before you merge.
Shared conventions
Same patterns enforced everywhere. What one team calls "how things are done" becomes the standard for the rest.
Pricing
Simple pricing. Scale when you're ready.
Pro
$50/mo base + $25 per seat (min 3 seats)
- Min 3 seats, unlimited max
- Unlimited projects
- 100 PR reviews/seat/month
- Health + Architecture + Conventions
Enterprise
Full data sovereignty. Your infrastructure.
- Unlimited seats and PR reviews
- BYOK: OpenAI, Anthropic, Azure, Bedrock, self-hosted
- Private deployment (Docker or Kubernetes)
- SSO, SLA, and dedicated support
The codebase doesn't have to rot.
Connect a repo. See the score. Track the drift.
Plans start at $50/mo.