Steph Mark

Huge day at @merge_api! Introducing GATEWAY 🎉 Smarter LLM routing, auto fallback, built-in cost guardrails, and enterprise-grade security… all in one API. Our bread and butter. Get $10 in free LLM credits to try it now! So proud of this team.

I've taken the last two products at Merge from 0→1. Nothing matters more in that phase than your first customers. The ones who give you raw feedback, push you through every edge case, and help you build something that actually works. @trywindmill was one of those customers for Gateway. They migrated from OpenRouter in a single PR and are now saving over $10K/month on LLM spend. Their engineer put it plainly: "Merge Gateway offers better LLM routing, cost control, and observability than OpenRouter." What they're running: - Intelligent routing across Claude Sonnet 4.6, Haiku 4.5, Kimi K2.5, and GLM-5 - Spend caps and alerts so finance stops getting surprised - Centralized governance and request-level logs Seeing a customer go from launch to scaling and saving real money. That's the most rewarding part of building. I recommend checking out how Windmill took a bet on us and started seeing the dividends immediately: merge.dev/case-studies/w…

@yashgogri1 Yessss!!

I've been thinking a lot about why great products get abandoned. It's almost never the product. It's the gap between signup and first value. So I rebuilt Agent Handler's onboarding around one idea: get to your first tool call as fast as possible. Pick your use case. Follow the path. Minutes, not hours. Check it out here: ah.merge.dev

We support zero data retention with @xAI so you don't have to. It’s live now and you can try it free for the next 7 days. → gateway.merge.dev/signup

You can't build production agents without a multi-model ecosystem: model routing, cost management, and governance included. OpenAI models finally landing on AWS Bedrock proves it: users want access to the best models without rebuilding their infrastructure. That's exactly what we built Merge Gateway to solve. One Unified API to offer all models from folks like OpenAI, Anthropic, Bedrock, Google, and more.

GPT-5.5 is available via Merge Gateway. Start routing to @OpenAI's latest model that handles complex, multi-step tasks with greater accuracy in coding, analysis, and writing, leading it to rank #1 on the Artificial Analysis Intelligence Index across 147 models. Start testing it today → merge.dev/gateway

I've been using @ahrefs in Claude Code over the past few weeks and it's completely changed how I work on SEO and AEO. In literal seconds I can identify visibility gaps with competitors, prioritize keyword and prompt opportunities, and spot pages that need to be refreshed.

Welcome to the canvas, @merge_api! n8n users can access 20+ new connections through the Merge Agent Handler node. 🤖

Super hype to use Claude Opus 4.7, but also wanted to share some advice for folks who may be using some of the older models soon to be deprecated. Updating models may be a quick code change, but the reality is that it's more than that. You need to run evals, make trade-offs around cost, factor in changes in latency, and more. There's a lot that goes into transitions. Here's how Merge Gateway can help: - Extend model support timelines via offering the LLM via multiple vendors - Automatic routing when a model is deprecated, no manual intervention - Intelligent routing to seamlessly add new models into your agents without manual intervention - Access to top models across all major providers in one place 𝗢𝗻𝗲 𝗔𝗣𝗜, 𝗻𝗼 𝗻𝗲𝘄 𝘃𝗲𝗻𝗱𝗼𝗿𝘀, 𝗰𝗼𝗻𝘁𝗿𝗮𝗰𝘁𝘀, 𝗼𝗿 𝗯𝗶𝗹𝗹𝗶𝗻𝗴 𝘀𝘆𝘀𝘁𝗲𝗺𝘀

When the barrier to ship "good" product is lower than ever, customer experience and reliability are even more important. It's a requirement to build in fallback and intelligent LLM routing for any customer facing AI features - you can't have your customers using broken product. Instead of building it yourself, leverage Merge Gateway to handle it all for you: merge.dev/gateway

Let me show you how easy it is to implement smart routing in your LLM-based projects. Smart routing will select the best model based on the difficulty of the request. This is honestly one of the easiest things you could do to make your app 10x better.

@andruyeung You should meet @GilFeig and @shensi they just launched the sickest LLM routing product @merge_api !!!

In SF next week - who should I meet? Also, who wants to get a workout/cold plunge in?

Merge's superpower has always been making complicated infrastructure easy. They've done it again!

@shensi @merge_api LFG!!!

We just built the #1 tool every AI team wants. Introducing @merge_api Gateway: LLM routing, fallback, cost guardrails, and security, all in one place. Everyone gets $10 free LLM usage on us to try it out. RT+ comment “Gateway” → we’ll double your credits.

@n0w00j can men handle multiple worktrees??

@yashgogri1 @dharmesh @merge_api oooo I'm excited for this one!!! 👀

And for anyone facing disruptions to key AI workflows or customer experiences, @merge_api is cooking up something perfect for you. This happens too often to not have a solution like ours backing you up Coming soon, T-3 weeks.

Current Status: Dealing with the emotional roller coaster that is the Anthropic API right now (they're having an owie). On the one hand, I'm in a state of flow and trying to finish up some work for a demo video I want to record a screencast for. On the other hand, it's draining to keep retrying things 5 times in order to get it to work. Just creeping along. My best wishes to the Anthropic team. Never fun to be going through system issues -- especially on a Friday afternoon.

we've launched 100+ spectacular connectors + we’re launching several every week, but we’re taking a more methodical approach to building each. I broke down our team’s process here: hubs.li/Q046y7YM0

If you hate MCP, you don't understand MCP. You have a bad setup/harness and blame it on MCP, which is nothing more than an extremely thin wrapper. The argument that it's slow is rooted in poor implementations that load up thousands of tools and eat your entire context window. But that's not necessary and it's not how the best AI companies are using MCP in practice. MCP is great because it's a lightweight standardization similar to REST and API protocols. It gives the world the ability to provide a minimal, single interface into their products. It then plugs nicely into all LLMs. MCP or not, we need that. Good MCP has 2 tools. Search tools, and run tool. You don't need anything more than that. You expose millions of tools to your agent, and they run quickly and eat close to zero context. CLI tools are great too. But they're heavier to produce, heavier to set up, require running agents to have CLI access, and give up a ton of security guarantees. MCP was invented when CLIs already existed for a reason. Many reasons. We're not suddenly enlightened by CLIs. Nothing has changed.