Skip to content

AIagentswritethecode.
unerrenforcesthearchitecture.

The missing backend for AI coding agents — a living knowledge graph that turns every AI suggestion into architecture-safe code.

unerr
B+84/100· 14,832 entities
TypeScriptReactNode.jsPrismaSynced 2m ago
B+

Good

🔥 3w improving

Solid foundation with a few structural blind spots

1 high3 medium2 auto-fixed

Top Issues

View all 42
3
Circular Dependency

Core modules create import cycles that confuse AI agents

12
Dead Code

Unreachable functions consuming 2,400+ lines

8
High Fan-In

Modules with 20+ dependents — high blast radius on change

Intelligence
82% confidence

14,832

entities indexed

8communities
Explore call graph
Guardrails
Enforcing

5

active rules

3patterns
Manage rules
Pipeline
completed

12s

Last run 3m ago

100%success
View pipeline runs
PR Reviews
1 issues found

8

PRs reviewed

1blocked
View reviews
Documentation
75% verified

12

docs tracked

9verified
3stale
View doc health
Atlas

8

feature areas mapped

8domain terms
Explore atlas

Architecture Composition

Full graph
Logic 42%
Plumbing 31%
Utility 27%
12 dead code entities detected — clean up to reduce bundle size
0K+
entities indexed
0
risk types
< 0s
first audit
< 0ms
query latency

Every AI coding agent today is a brilliant intern with amnesia.

Without unerr

  • No codebase memory
  • Ignores existing patterns
  • No blast radius awareness
  • Can’t rewind AI changes
  • Stale context, wrong answers

With unerr

  • Living knowledge graph
  • Pattern enforcement
  • Blast radius maps
  • Full audit trail & rewind
  • Verified docs via MCP

“AI co-authored code contains 1.7× more major issues, 75% more misconfigurations, and 2.74× higher security vulnerabilities.”
— Based on analysis of 470+ open-source GitHub PRs

Zero to protected in 60 seconds.

One command. Full graph. Zero config.

1
Step 1 — One command

Connect

Install the GitHub App or run npx @autorail/unerr setup. One command detects your IDE, authenticates, and configures MCP.

$ npx @autorail/unerr setup
✓ Detected Cursor IDE
✓ Authenticated via GitHub
2
Step 2 — Automatic

Analyze

The knowledge graph builds automatically. Every file, function, and relationship — indexed with semantic embeddings and business justifications.

✓ MCP server configured at localhost:9315
✓ Indexing 3 repositories...
3
Step 3 — Always-on

Supervise

Your AI agents read the graph via MCP at localhost:9315. 80% of queries resolve locally in <5ms. The rest route to cloud.

✓ Health audit complete — Grade: A (92/100)
unerr is now supervising your AI agents.

What your AI agents are missing.

One knowledge graph that remembers what your agent forgets.

Living Knowledge Graph

Every file, function, and relationship — indexed, connected, and always current.

Works With Every Agent

One MCP server. Full codebase context in any agent.

Prompt
"Fix the login timeout bug"
⚡ unerrcontext injected
3 dependency chains · 4,200 tokens saved
Auth boundary rules applied
CursorClaude CodeVS CodeWindsurf

Codebase Health Report

Dead code, coupling, architectural drift — surfaced in 60 seconds.

A
92 / 100
Battle-tested — minimal risks
Type Safety
96 Pass
Dead Code
88 Pass
Coupling
74 Warn

Blast Radius PR Review

Every PR checked against the graph. Downstream impact before merge.

// checkout/payment.ts
14- await db.query('SELECT * FROM users')
14+ await userService.findById(id)
15 const invoice = createInvoice(user)
16 // 3 downstream callers affected
Safe to merge2 features affected · 0 violations · 12 entities in blast radius

Architectural Rules Engine

Define boundaries in plain English. Auto-detected from history, enforced on every commit.

{
pattern: "no-db-in-ui",
scope: "components/**",
rule: "No direct database imports in UI layer"
}
3 violations foundAuto-detected from 6 months of commit history

Prompt Ledger & Rewind

Full audit trail of every AI change. Rewind with one command.

a3f2c1dRefactor auth
Claude · 4 files
c1d8f3eUpdate schemaRewind
GPT-4o · 7 files
d9a2b6cFix tests
Claude · 2 files

Infrastructure-grade. Developer-priced.

Free forever for open source. One simple plan for everything else.

OSS Guardian
$0/ forever

Full unerr Pro for verified open-source projects. Because maintainers deserve a tech lead, not another AI bot.

Full knowledge graph
Health audit & grades
MCP server integration
AI PR review
Pro
$20/ month

Everything you need for safe AI velocity on private codebases.

Full knowledge graph
Blast radius analysis
MCP server integration
13-type health audit
Architectural rules
Prompt ledger & rewind
AI PR review
Unlimited private repos

Teams from $45/mo (3 seats × $15) · See all plans →

Your AI agents need architecture.
Give them unerr.

50K+
entities indexed
13
risk types
< 60s
first audit

Free for 7 days · No credit card · Setup in 60 seconds