FlurryPORT

About FlurryPORT

FlurryPORT began as a webhook capture-and-replay service and grew into an action broker for AI agents, and both halves rest on the same primitive: a capture is the exact bytes a sender sent, stored unmodified and sendable again. As a webhook tool it records every header and every byte as the provider transmitted it, auto-detects Stripe, GitHub, Shopify, Slack and Twilio payloads, and verifies Stripe, GitHub, Svix and Square signatures before a single byte is stored, so a spoofed request gets a 401 and a flagged capture instead of being silently swallowed. Replay works three ways — one request, a batch selection, or sequential replay that reproduces a production bug in order — and each developer gets isolated capture endpoints with scoped replay targets, which removes the shared-tunnel collisions that come with ngrok. One capture can be forwarded to several destinations and reshaped into each one's schema on the way out. The agent surface is built on top: you install a recipe and the agent gets one small typed action it is allowed to request — file a GitHub issue, post to Slack or Discord — while the credential stays in the secret store and is attached server-side at delivery, so the model never holds it, and what comes back is the service's own response rather than the agent's account of it. Rooms give several people and several agents one ordered, signed record. Payloads are envelope-encrypted with AES-256-GCM and automatic key rotation, and recipes install through an MCP server.

Key Features

Tags