Code intelligence for AI coding agents

If your agent writes
bad code. This catches it.

Scout finds security, performance, correctness, and architecture issues across your codebase. Then it maps repository relationships so your coding agent can make changes with the right context.

View on GitHub
$npm install -g @aykutss/scout

One command. Works with Claude Code, Codex, Cursor, and other coding agents.

repository contextindexed
code structureimportscallsscout flow
01Scan the repository

Rules, AST analysis, dependencies, history.

02Map the relationships

Imports, calls, symbols, exports.

03Feed the agent context

Structured findings and targeted code.

Two jobs. One system.

Catch the defect. Show the surrounding code.

A finding without context leaves the agent guessing. Scout combines repository-wide analysis with a structural context tree, so the agent sees both the problem and how the codebase fits together.

01

Find what ordinary linting misses

scout scan1 finding shown
warningnext.local-server-boundary

Client component reaches server-only code.

Scout follows the local import chain instead of checking a single file in isolation.

client componentlocal moduleserver-only code
category: bugconfidence: rule
02

Give the agent structured scan output

scout scan
JSON output
score95
findings1
errors0
warnings1
{  "summary": {"total": 1, "score": 95},  "findings": [{    "analyzer": "next-hint",    "rule_id": "next.local-server-boundary",    "severity": "warning",    "category": "bug",    "confidence": "rule",    "file": "app/client.tsx",    "line": 4,    "message": "Client component imports local server-only code through \"./db.server\"; reaches app/db.server.ts.",    "fix": "Keep server-only code in a Server Component, Server Action, route handler, or API layer."  }]}

Piped output is structured for coding agents.

What Scout checks

More than another linter.

Scout wraps proven analyzers and adds custom Semgrep and tree-sitter rules for the gaps between files, frameworks, and layers.

01

Security

Find exposed secrets, unsafe data flow, missing guards, vulnerable dependencies, and framework-specific risks.

02

Performance

Catch costly render patterns, repeated work, inefficient collection operations, and avoidable runtime allocation.

03

Correctness

Surface broken async flows, lifecycle mistakes, unsafe boundaries, and defects that ordinary linting misses.

04

Architecture

Inspect imports and cross-file relationships for cycles, layer violations, oversized components, and missing registrations.

JavaScript ecosystem

Framework-aware from the start.

Scout detects the project it is running in and enables relevant analyzers instead of applying every rule everywhere.

01JavaScript02TypeScript03React04Next.js05Svelte06Node.js07Express08NestJS09Vite10TanStack11React Native12Bun

From install to useful output

Built for the way coding agents work.

01

Install once

The npm package installs the Scout command and its skill for detected coding agents.

$npm install -g @aykutss/scout
02

Run from the agent

Use one skill to scan the repository and another to build its complete structural context tree.

/scoutscan
/scout-contextcontext tree
03

Use it anywhere

Skills drive coding agents. The same capabilities are available directly from the terminal.

$ scout scan$ scout context

Give your coding agent a second set of eyes.

If it writes bad code, Scout should catch it.

View on GitHub
$npm install -g @aykutss/scout