AppFactor

🚀 We are on the Product Hunt today! Now you can connect any API to any AI agent with MCP Bridge. Help us climb to the top: ⬆️ https://lnkd.in/dfpyHwzT

My friends from AppFactor launched MCP Bridge on Product Hunt today! 🚀 MCP is the new standard for connecting AI agents to tools and data. Setting it up usually means writing a custom MCP server per API. Their MCP Bridge skips all that - just point it at any REST, GraphQL, SOAP, or gRPC API and you get agent-ready tools. Now you connect any API to any AI agent. ⬆️ Support them with an upvote here: https://lnkd.in/da6q9hgE

MCP Bridge is live on Product Hunt today 🚀 Every platform team we talked to was building a stack of custom MCP adapters, one per service. That integration tax is the actual blocker to agentic AI in enterprises. MCP Bridge is the layer that removes it. APIs were built for human developers, not LLM agents. 🤖 This is why we bulit MCP Bridge, a self-hosted Rust platform that turns any REST, GraphQL, SOAP, or gRPC API into agent-ready MCP tools. One layer, no custom adapters, no data leaving your network. ⬆️ Upvote us on PH, link is under the comments. Our full demo by Keith Neilson:

The Developers conference poster agenda. Join the biggest technical summer event in Cluj for programmers. www.the-developers.ro Daniel Costea Vlad Mihalcea - CTO @ Hypersistence Cornel Stefanache - CTO @ AscentCore Gabriel Ciuloaica- CTO @ AppFactor Doru Șupeală - Managing Partner @ Hacking Work Tudor Damian - Cloud & Cybersecurity Advisor @ D3 Cyber Gabriel Ciuloaica - CTO & co-founder @ AppFactor Denis Salanta - Delivery Manager @ Banca Transilvania Roland Szabo - LLM Consultant @ Gradient Ascent Andrei Olar - PhD candidate @ Computer Science at Babeș-Bolyai University Andreea Giurgiu - Cloud Solution Arhitect in Data & AI (Internship) @ Microsoft Alexandru-Catalin Dinis - CTO @ BMW TechWorks Romania Ovidiu Mățan - Founder @ Today Software Magazine Denisa-Stefania Todoran - Java Software Developer @ msg Radu Jinga - Executive Vice President @ msg Adrian Rosian - Founder and owner @ WhiteCity Code Voicu Oprean - CEO and founder @ AROBS Group Tudor Ciuleanu (Tudi) - Founder and CEO RebelDot Horea Avram - Art Historian

Most MCP demos work. Few survive production. The difference: auth, rate limits, retries, observability. Today's MCP Bridge blog walks through how we built them: 6 auth methods (including Amazon Web Services (AWS) Cognito SRP), per-tool and per-tenant rate limits, jittered exponential backoff, active health probes, OpenTelemetry export. Read the features breakdown: https://lnkd.in/d2dUDgeP Stay tuned for the Product Hunt launch tomorrow 🚀

📢 I'll be speaking in a panel discussion next week at #southsummit26 🪧 "From Assistant to Operator: How Agentic AI Is Rewiring Business" 👋 See you there and happy to catchup if any of my connections are planning to attend.

AI Enrichment is live in MCP Bridge! 🌁 Most enterprise APIs were built before agents existed. Names like 𝘨𝘦𝘵𝘗𝘳𝘰𝘤𝘐𝘯𝘧𝘰3() and descriptions like "see docs" don't tell an LLM anything. The fix: read the response shape, not the function name. From WSDLs, OpenAPI, or a captured sample call. The LLM renames 𝘨𝘦𝘵𝘗𝘳𝘰𝘤𝘐𝘯𝘧𝘰3 to 𝘨𝘦𝘵_𝘤𝘶𝘴𝘵𝘰𝘮𝘦𝘳_𝘢𝘤𝘤𝘰𝘶𝘯𝘵_𝘥𝘦𝘵𝘢𝘪𝘭𝘴 for the discovery layer. Original schema preserved. Agent still calls the real API. Discovery accuracy on one customer's SOAP service: from 41% to 89%. 🔗 Read the full feature description: https://lnkd.in/dHmAhYHJ

Today's drop is Code Mode feature on MCP Bridge. The problem we kept seeing: tool definitions eat the context window. ~250 tokens per definition, and a real enterprise API has 100+ operations. Before the agent has done anything, the prompt is 50,000 tokens of schema. 🤯 What we shipped, instead of loading every tool definition, the agent gets 3 meta-tools: 1. search_tools: find the right tool by describing what you need 2. get_tool_schema: pull a specific tool's schema on demand 3. execute_code: run multi-step tool calls in a sandboxed JS runtime Net result on real enterprise API surfaces: ~98% context window reduction. The model only sees the tools it picks, only when it needs them. As a bonus, loops and conditionals live in execute_code, 800 customer lookups doesn't mean 800 inference round trips. Cloudflare and Anthropic have both written about Code Mode. What's new: it's built into the platform that already translates REST, GraphQL, SOAP, and gRPC into MCP. You get Code Mode without giving up multi-protocol coverage. Full writeup with benchmarks & early access under comments. Launching MCP Bridge on Product Hunt this Thursday 🚀

We started building MCP Bridge a few months ago because our customers kept hitting the same wall: their APIs are a mix of REST, GraphQL Foundation, SOAP, and gRPC.  🧩 Every enterprise agent POC we've seen hits the same wall, the demo works against a clean OpenAPI spec under The Linux Foundation, then someone asks "can it talk to the system the business actually runs on?" and the answer is a six-month integration project. So we built for this reality. Allowing any existing system and API to be reimagined as a high quality set of Model Context Protocol tools, with security and governance thats expected in production. Rust core, self-hosted, translates all four protocols into MCP. One layer, no per-service adapters. The "any system, any API" isn't marketing. It's the whole reason MCP Bridge exists. REST is the easy case. The hard cases are where agents go to die in production: SOAP services with WSDLs from 2009, GraphQL endpoints that return half the database, mainframe gateways behind a screen-scraping layer. This week we'll share how it actually works. Launching MCP Bridge on Product Hunt this Thursday! 🚀