MMRequest

Multimodal HTTP client. Drop an image and it auto-encodes to base64 in your API request body. For vision LLM developers.

🚀 MMRequest — the HTTP client built for multimodal LLM APIs. THE PROBLEM Every vision/audio LLM API (GPT-4o, Claude, Gemini, vLLM, Llama, …) wants your image or audio as a base64 string buried inside the JSON body. So today you end up uploading the file to some random website just to copy its base64, pasting a 300KB blob into Postman, and redoing the whole thing every time the file changes. It's slow, messy, and easy to get wrong. THE FIX Drop an image or audio file → MMRequest auto-encodes it to base64 (in your browser, no upload) and injects it into your request through a clean {{base64Image}} / {{base64Audio}} placeholder. Your JSON body stays readable; the heavy base64 is substituted only at Send time. Build, send, and iterate on multimodal requests the way a normal HTTP client never let you. ━━━━━━━━━━━━━━━━━━━━━━━━━━━ ✨ What's New in v0.0.3 ✨ 🎨 Themes — 10 hand-crafted color themes (7 dark + 3 light: Midnight, Ocean, Forest, Aurora, Crimson, Nord, Sunset, Daylight, Paper, Mint). Recolors the whole app, not just the accent. 🌗 Sync with system — auto day/night switching that follows your OS appearance, or pick a theme manually. 🗂️ Smarter tabs — multi-session tabs with a right-click menu (New / Close / Close Others / Close All) and unsaved-change indicators (🟠 editing · 🟢 saved) shown on tabs AND saved requests. 📦 Postman compatible — export any request as a Postman Collection (v2.1) and import Postman collections too. Real base64 media is embedded so the request runs anywhere. 🌊 Streaming responses — a new Stream tab merges chunked NDJSON / SSE output (including reasoning tokens) into one clean, combined JSON. 🌲 Collapsible JSON viewer — Pretty view folds each field; long strings collapse with "…expand" so base64 no longer floods the screen. 🔢 Line numbers in the Raw response view. 🧰 Collections — drag & drop to reorder or move requests between collections, pin favorites, rename / duplicate / copy via the ⋯ menu, and search across all requests. 🧾 Grouped console — each request/response cycle folds into a single expandable log entry. ⌨️ Shortcuts — Cmd/Ctrl + S to save a request, Cmd/Ctrl + E to rename. 🐛 Fixes — dropped/pasted media now always attaches to the active tab, plus lots of UI polish. ━━━━━━━━━━━━━━━━━━━━━━━━━━━ MMRequest is a Chrome extension HTTP client built for one specific workflow that existing tools don't automate well: calling multimodal LLM APIs (GPT-4V, Claude vision, vLLM, Llama vision, etc.) that need a base64-encoded image inside the JSON body. If you've ever: - Uploaded an image to a 3rd-party site just to get its base64 string - Copy-pasted a 300KB base64 string into Postman's JSON body - Forgotten to update the {{image}} placeholder before re-sending - Debugged the "wrong Content-Type" issue with multimodal endpoints … this extension is for you. ━━━ HOW IT WORKS ━━━ 1. Click the toolbar icon → MMRequest opens in a full-screen tab 2. Set your API URL (supports variables: http://{{host}}:4017/v1/chat/completions) 3. Drop an image into the right-side dropzone (or paste with Ctrl+V) 4. The image is auto-base64-encoded and bound to {{base64Image}} 5. Your JSON body keeps the placeholder for readability — substitution happens only at Send time 6. Click Send → see the response below (Pretty / Raw / Stream / Headers / Console tabs) ━━━ FEATURES ━━━ CORE • Drag & drop, file picker, and clipboard paste for images • Auto base64 encoding via FileReader (no network round-trip) • Multiple images supported ({{base64Image1}}, {{base64Image2}}, ...) • Variable system with {{host}}, {{api_key}}, etc. • Live "Resolved URL" preview showing the actual request URL REQUEST EDITING • HTTP methods: POST / GET / PUT / DELETE / PATCH • Custom + auto-generated headers (toggle on/off) • Body modes: Raw JSON with syntax-aware textarea • Authentication: None / Bearer Token / Basic Auth / API Key (header or query) • Variables tab with image variable auto-binding RESPONSE VIEWING • Status, duration, size, token usage (vLLM `usage` field) • Pretty (JSON formatted), Raw (line numbers), Stream (streamed merge), Headers • Console tab with full request/response logs and "Failed to fetch" diagnostics • Copy response / Copy cURL / Copy body buttons WORKFLOW • Request history (last 50, click to restore) • Collections / Favorites with tree structure • Postman Collection import/export, cURL export • Cmd/Ctrl + Enter to send • Settings persist across sessions (localStorage) ━━━ PRIVACY ━━━ MMRequest is privacy-first by design. • All request/response data stays in your browser. We never see your URLs, headers, bodies, or responses. • Images you drop are encoded in your browser via FileReader — they never touch our servers. • The only data we collect is your Google email (when you choose to sign in for Early Adopter benefits) and basic sign-in event timestamps. • Full privacy policy: https://mm-request.vercel.app/privacy ━━━ EARLY ADOPTER BENEFIT ━━━ We are currently in beta. Users who sign in with their Google account before our public launch will retain free access to all features after we introduce paid plans. ━━━ PERMISSIONS WHY ━━━ • storage: Saves your URL, headers, variables, collections, and history locally so they persist across sessions. • host_permissions <all_urls>: MMRequest is an HTTP client where you enter your own API endpoints. We cannot know in advance which hosts you will call. No automatic requests are ever made; every request is triggered by you clicking Send. • identity / identity.email: Reads your Google email when you click "Get started with Google" to track Early Adopter eligibility. Not used for anything else. ━━━ FEEDBACK ━━━ This is an actively developed beta. Send feedback or feature requests: rex.leto.addib@gmail.com Website: https://mm-request.vercel.app