Manta Action Kit - AI Developer Toolkit & Agent Browser Bridge

Secure access to your authenticated APIs for Claude Code & Codex. Record business flows via a sandbox proxy — cookies stay private.

🔐 Give your AI agents secure access to your real, logged-in APIs — without ever handing over your credentials. Tools like Claude Code and Cursor are great at writing code, but they can't safely call your authenticated business APIs. Manta Action Kit fixes that: you record the real API flow once, then let your agent call those endpoints through a controlled sandbox proxy — while your cookies and tokens stay locked inside your browser. How it works: • Record: You actively capture a real business API flow (fetch/XHR) in your browser — requests and responses included. • Sandbox Proxy: Your agent sends only method/URL/body. The extension injects your session cookies at the trust boundary and forwards the call, so credentials are NEVER exposed to the AI. • Stay in control: human-in-the-loop confirmation on every agent call, a per-tool kill switch, an SSRF guard that refuses loopback/private hosts, and a full audit log (cookie names logged, values never) — you decide what the agent can reach. Built as an AI Developer Toolkit, bridged to your agent over MCP. (Coming Soon: DOM Simplifier, Authentication State Injector, and more!)