POST /orders order-service Validate customer order-service cross-repo Check payment payment-svc Save order → DB Kafka emit order.created Notify customer notif-service same repo cross-repo 0.95 ✓
AI-Augmented Semantic Logic Mapper

See the Thread,
not the Hairball.

In a distributed architecture, the source of truth for any business process is scattered across a dozen repos. LogicLoom traces the exact path of a business event — from API call to database write — across every service, automatically.

REGISTER INTEREST See how it works
4 languages
Java · Go · Python · TypeScript
<30s
Incremental index per PR
MIT core
Open source, self-hostable
The Problem

Every team is flying blind through their own architecture.

When a PaymentFailed event fires in production, who knows every service it touches? When a new engineer joins, how long before they can make a change without fear? When the Confluence diagram was last updated — does anyone actually know?

Standard tools show you everything. LogicLoom shows you the specific thread — the exact path of one business event, across every repository, labeled in plain English.

team-standup.slack
AK
What exactly happens when a payment fails? Need to trace it for the incident.
LL
$ loom trace "POST /payments/charge" --on-fail
Step 1Validate card via Stripe SDK
Step 2Write FAILED status → orders table
Step 3Emit payment.failed → Kafka
Step 4notification-service → Email + SMS
AK
That took 8 seconds. Our last RCA took 3 hours manually.
Without LogicLoom

The Hairball. Everything. Noise.

You see every node, every edge, the full system — a dense, undifferentiated graph that answers nothing.

  • RCA takes hours of manual repo-hopping
  • New engineers take 3–6 months to navigate safely
  • Confluence diagrams drift from reality within weeks
  • Impact of a change is a gut-feel guess
  • Architecture policy violations discovered in production
  • Compliance audits require weeks of manual documentation
With LogicLoom

The Thread. One path. Perfect clarity.

Select an entry point, select an exit point. LogicLoom illuminates only the path between them — labeled in plain English, confidence-scored at every edge.

  • Trace any business event in seconds from the CLI
  • New engineers productive in days, not months
  • Graph always in sync — auto-updated on every commit
  • Blast radius analysis before every deploy
  • Policy-as-Code gates catch violations in PR review
  • Compliance reports generated automatically from the graph
Core Capabilities

Built for real distributed systems.

Not toy examples. LogicLoom is engineered for the complexity you actually face: dynamic Kafka topics, Spring AOP proxies, runtime-constructed URLs, and 50 repos in the same workspace.

Open Source
Cross-Repo Stitching
Automatically connects EmitEvent("order.created") in one repo to @KafkaListener in another. Five stitching strategies with confidence scoring — from spec-match (0.95) to fuzzy-match (0.60).
Open Source
Semantic Labeling
Replace private void validateUser(User u) with "Step 1: Business Validation." AI-generated labels (local Mistral or cloud Claude), human corrections as ground truth, audit trail on every change.
Open Source
Confidence Scoring
Every edge carries a confidence score. Low-confidence edges render as dashed lines. Uncertainty is always visible — wrong diagrams are worse than no diagrams.
Pro
Policy-as-Code
Define architectural rules in YAML — "adapters must never call the database directly" — and enforce them automatically in every PR. Violations block merge at the CI gate.
Pro
PR Logic Diagrams
Every pull request automatically gets a "Logic Change Diagram" comment showing what architectural paths changed. Before-and-after diff view. Deep-link to the full interactive canvas.
Enterprise
Drift Detection
Architectural drift — the gap between your diagrams and reality — detected automatically. Compliance reports generated from the live graph for SOC2 and ISO 27001 audits.
How It Works

From zero to a traced logic flow in under 5 minutes.

01
Point it at your repos
One config file. List your repos and their languages. No JVM, no Docker required — loom-agent is a single Go binary that runs anywhere.
# loom.yaml repos: - name: order-service url: git@github.com:acme/order-service.git - name: payment-service url: git@github.com:acme/payment-service.git
02
The Crawler & Stitcher run
Tree-sitter parses your code into ASTs. The Stitcher connects cross-repo calls — HTTP, Kafka, gRPC — using spec-matching, service registry lookups, and fuzzy-matching, each with a confidence score.
$ loom index ✓ Parsed 847 files across 4 repos (12.3s) ✓ Extracted 2,341 boundary nodes ✓ Stitched 189 cross-repo edges → 142 spec-match (confidence 0.95) → 31 registry (confidence 0.85) → 16 fuzzy (confidence 0.60) ✓ Labeling 94 significant nodes…
03
Ask it anything about your architecture
Trace the path of a business event from entry to exit. See which services it touches, what the AI labeled each step, and how confident the stitcher is about each edge.
$ loom trace "POST /payments/charge" --to "orders_table" Step 1 Accept payment request [order-service] ↓ CALLS → confidence 0.95 (spec-match) Step 2 Validate customer credit score [order-service] ↓ CALLS → confidence 0.87 (registry) ← cross-repo Step 3 Charge card via Stripe [payment-service] ↓ WRITES_TO → confidence 1.0 (runtime-confirmed) Step 4 Save order to database [order-service]
04
Embed it in your team's workflow
VS Code extension shows the logic flow for the file you're editing. The GitHub App posts a logic change diagram on every PR. Policy rules catch architectural violations before code merges.
# In your PR — automatically posted by LogicLoom 🔍 Logic Change Analysis + NEW: order-service → notification-service (Kafka) ~ CHANGED: PaymentService#processPayment label updated ✗ POLICY VIOLATION: adapter-layer accessed DB directly → PaymentAdapter.java:87 (severity: ERROR)
Early Feedback

Engineers who live with this problem every day.

We have 60+ microservices. When something breaks, three engineers spend half a day just figuring out the blast radius. A tool that traces this automatically would change how we do incident response.
SR
Staff Engineer
Series B fintech, 200 engineers
The architectural drift problem is real and it compounds. We had a service calling a database it wasn't supposed to touch for eight months before anyone noticed. That kind of thing would be caught on day one with policy gates.
MK
Principal Engineer
Enterprise SaaS, 500+ engineers
Onboarding is our biggest pain. New engineers take months to feel safe making changes. If I could give someone a live, labeled map of the system on their first day, I'd cut that time in half.
TP
VP Engineering
Growth-stage e-commerce, 80 engineers
Early Access

Be first to map
your architecture.

We're onboarding early access teams now. Register your interest and we'll reach out when LogicLoom is ready for your stack.

✓
You're on the list. We'll reach out when early access opens — usually within 2–4 weeks of submission.

No spam. No marketing drip. Just a direct ping when we're ready.

340+
Engineers on waitlist
18
Early access teams
4
Language parsers
MIT
Core license