Kampala
A tool that watches an app's network traffic while you use it normally, then automatically generates a working API from what it observed — turning a website, mobile app, or old internal system with no API into one you can actually script against, without
🔗 Visit Kampala
Description
Lots of important business systems — a freight carrier's booking portal, an old insurance platform, an internal ERP screen — were never built with an API a developer could call directly, which normally means either paying for expensive custom integration work or building fragile browser automation that breaks the moment a button moves. Kampala takes a different approach: you just use the app normally once while it watches in the background, and it turns everything it saw — every request, every login step — into a proper, typed API you can call directly going forward.
Kampala, by Zatanna (YC W26), is a man-in-the-middle proxy tool that captures an application's network activity during a 'capture session' — recording HTTP requests, WebSocket frames, and GraphQL queries across websites, mobile apps, or desktop apps — then converts that traffic into a structured 'surface': typed API endpoints with inferred authentication, rate-limit handling, and dependency mapping. It targets developers and startups in healthcare, logistics, hospitality, and other industries needing programmatic access to legacy back-office tools, ERP systems, freight carrier portals, or insurance platforms that don't expose their own API. It offers both an agent harness that builds the integration by prompting you through actions, and an MCP interface so a coding agent (like Claude Code) can use Kampala directly to turn an observed workflow into a script or hosted API. No pricing is published; the only options are booking a demo or a workflow review call.
💬 Our review
The short version: Kampala is a sharp, narrow tool for a developer who's stuck automating a legacy system with no API — an old ERP screen, a carrier's booking portal — and is tired of brittle browser-click automation breaking every time the target site changes something cosmetic.
The core idea, capturing real network traffic once and turning it into a typed, reusable API rather than replaying UI clicks, is a fundamentally more robust approach than typical browser automation or 'computer use' agents, which tend to break whenever a page's layout shifts even slightly. Targeting specifically the systems that never had a real API to begin with (freight carriers, insurance back-offices, internal ERPs) is a smart, underserved niche rather than trying to compete head-on with general-purpose scraping tools. The honest limitation: it's an early-stage (YC W26), enterprise-sales product with no public pricing, so you can't just sign up and try it — evaluating fit requires a demo call, and as with any reverse-engineering approach, results depend on how well-behaved the target application's traffic actually is.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Aucune tarification publique ; uniquement démo ou session de revue de workflow (30 min). Vise des cas d'usage entreprise (santé, logistique, hôtellerie, assurance).
Pros
Capture le trafic réel plutôt que de rejouer des clics UI — bien plus robuste face aux changements de design
Génère une API typée avec authentification et gestion de rate-limit inférées automatiquement
Cible spécifiquement les systèmes sans API existante (ERP, portails transporteurs, assurance)
Interface MCP pour laisser un agent de code (Claude Code) construire l'intégration directement
Cons
Aucune tarification publique, nécessite un appel commercial pour évaluer
Produit très jeune (YC W26), peu de recul d'usage
Efficacité dépend du comportement du trafic de l'application cible