model-context-protocol
(37)
Benjamin Mann and the Long Detour to a Safer Machine
He left a comfortable engineering career, changed course after one unsettling book, and helped build both GPT-3 and Anthropic. Now Ben Mann works where product ambition and AI caution keep sharing the same desk.
FiscalNote Turns Capitol Chaos Into a Search Bar
For teams buried under bills, rules and hearings, FiscalNote offers one disciplined place to watch government move. The coverage is broad, the workflow is practical, and the price still begins with a conversation.
Give Claude the Whole Assignment
Claude has grown from a careful chatbot into a workbench for writing, research, code and files. It rewards people who bring context, judgment and a job worth finishing - but its shifting usage limits still demand a little housekeeping.
Dust Wants Your AI Agents in the Group Chat
Dust turns scattered company context into shared AI work - useful for teams ready to build repeatable agents, provided they watch the credit meter and keep humans in the loop.
Daloopa Makes the Spreadsheet Sweat Less
The dullest hour in equity research is also the easiest to get wrong. Daloopa wants to give it back, one source-linked cell at a time.
Supernormal Turns the Meeting Hangover Into Finished Work
The former AI notetaker has grown hands, a desktop app, and a taste for client work. Its best trick begins after everyone leaves the call.
TravelPerk Grew Up. Your Expense Report Came Along.
The booking tool once known as TravelPerk now wants the whole awkward trip - approval, receipt, invoice and all. Here is where Perk earns its keep, where the fees bite, and who should put it on the shortlist.
CData Built the Boring Plumbing. Now 10,000 Companies Need It for AI.
The enterprise AI race has produced a very old problem: the useful data is scattered everywhere. CData spent a decade standardizing those connections, and its unglamorous specialty has become the layer everyone suddenly needs.
Dylan Etkin Measures the Moment Software Becomes Real
After helping Jira grow and watching Bitbucket scale, Dylan Etkin found the overlooked event that tells an engineering team whether its work matters: the deployment. Now he is applying the same practical instinct to AI skills, access and organizational memory.
Karim Rahme Is Building the Plumbing AI Agents Keep Tripping Over
He sold concert tickets in Abu Dhabi, graduated from NYU, got rejected by Y Combinator, tried again, and turned a mess of OAuth flows into Metorial. Karim "Wen" Rahme's useful obsession is the unglamorous layer between an AI agent and the real world.
Achintya Ashok Is Teaching AI Agents Which Tools to Pick
After building systems for currency traders, streaming audiences and crypto customers, the OneGrep co-founder is tackling a quieter AI problem: giving agents the judgment to use their tools well.
Andrew Sy Is Teaching Software to Read the Room
After building inference systems at Scale AI and security software at Opal, the math-competition alumnus is betting that better growth technology begins with a modest skill: knowing when not to interrupt.
Runlayer Wants Your AI Agents Working - Just Not Running Wild
Employees were connecting AI to company systems faster than security teams could count the connections. Runlayer turned that mess into a control plane - and persuaded Gusto, Jane, Homebase and other large companies to let the agents in.
The 'Where to Buy' Button Grew Up and Learned to Talk to Robots
PriceSpider spent 20 years turning a little 'Where to Buy' button into a business. Now, rebranded as Wayvia, it is wagering that the next shopper won't be a person at all - it will be an AI agent doing the buying.
Aman Singla Keeps Building the Missing Layer
From wireless chips to cloud-managed homes to governed AI agents, Aman Singla has spent three decades working on the infrastructure between a promising idea and the messy systems it must survive.
Wojciech Blaszak
Wojciech Blaszak is the co-founder and CEO of Golf (YC X25), a San Francisco company building a security and governance control plane for the Model Context Protocol (MCP), the layer where AI agents connect to company data. He started his first business at 14, grew it to around $200k in revenue, and left university after a single semester to build full-time with his longtime collaborator Antoni Gmitruk. Golf began as an open-source framework for shipping production MCP servers and grew into an enterprise product that discovers shadow AI, enforces per-tool policies, and produces compliance-ready audit trails.
Corsair
Corsair is an open-source integration layer for AI agents. Its TypeScript framework lets developers connect their agents and apps to hundreds of third-party services - Gmail, Slack, Stripe, Notion, GitHub and more - while Corsair handles the messy parts: OAuth, token refresh, webhook verification, rate limiting, permissions and multi-tenant credential isolation. Agents never touch raw API keys; Corsair resolves them internally and can require human approval before destructive actions. Founded in 2024 by Dev Jain and part of Y Combinator's Winter 2025 batch, the project is offered as a free self-hosted library, a hosted hub with a drop-in MCP URL, and a managed cloud SDK.
Golf
Golf is a Y Combinator (X25) startup building the security and governance layer for AI agents and Model Context Protocol (MCP) servers. It began as GolfMCP, an open-source Python framework for shipping production MCP servers without boilerplate, and grew into an enterprise control plane that discovers every AI agent and MCP connection in an organization (including shadow AI), enforces granular policies with sub-millisecond latency, and maintains compliance-ready audit trails. Founded by high-school friends Wojciech Blaszak and Antoni Gmitruk, who both dropped out of university to build it full-time.
Metorial
Metorial is agentic infrastructure that connects AI agents to more than 1,200 tools and data sources through the Model Context Protocol (MCP). Its serverless runtime hosts MCP servers that boot in under a second and hibernate when idle, while handling OAuth, permissions, RBAC/SSO, and full tracing so companies can ship customer-facing agents without building integration plumbing. Founded in 2025 by Karim Rahme and Tobias Herber, the open-source, Apache-2.0 project is part of Y Combinator's Fall 2025 batch and is often described as the 'Vercel for MCP.'
OneGrep
OneGrep (YC W24) is a New York-based developer-tools startup building AI agent infrastructure for DevOps. It began as a DevOps copilot that fine-tuned models on real-time telemetry to help engineers resolve incidents faster and cut observability costs, and has since expanded into tooling that lets AI agents discover, correlate, and safely operate the many SaaS and MCP-based tools engineering teams rely on. Founded by ex-Disney, Amazon, and Coinbase engineers Achintya Ashok and Brian Cripe, it went through Y Combinator's Winter 2024 batch.
Orthogonal
Orthogonal is a San Francisco startup building the discovery, orchestration, and payment layer for AI agents. Through a single SDK, REST API, MCP server, or CLI, agents can find, call, and pay per-use for 35+ third-party APIs - from data enrichment to search and scraping - with no subscriptions, no key management, and no sales calls. Founded in 2025 by Christian Pickett and Bera Sogut, the company raised a $4.3M seed round led by Pantera Capital and is part of Y Combinator's Winter 2026 batch.
Osmosis
Osmosis is a San Francisco reinforcement learning platform (YC W25) that helps companies fine-tune open-source models to beat foundation models on cost, latency, and reliability. Founded by Kasey Zhang and Andy Lyu, it lets teams train task-specific AI agents with multi-turn tool use over the Model Context Protocol, then export or self-host the weights they own.
Rowboat
Rowboat (YC S24) is a San Francisco and Bengaluru startup building open-source AI agent software. It began as a pre-trained customer-support LLM agent, grew into a widely starred open-source IDE for building multi-agent systems on the OpenAI Agents SDK, and is now 'your AI coworker, with memory' - a local-first desktop app that turns your emails, meetings, and notes into an editable knowledge graph so agents can actually get work done. Founded by the team behind Agara (acquired by Coinbase in 2021).
Vespper
Vespper (YC F24) builds infrastructure that lets AI agents produce real Microsoft Word documents. Its product is a docx MCP (Model Context Protocol) server that turns raw agent output into clean, tracked, review-ready .docx files - aimed at regulated, legal, and compliance workflows where the final artifact has to be a Word doc with tracked changes, not a chat window. The two-person, Tel Aviv-based team originally launched an open-source, 24/7 AI on-call engineer for incident triage before pivoting into agent document infrastructure.
Saoud Rizwan
Saoud Rizwan is the founder and CEO of Cline, an open-source AI coding agent that runs inside code editors like VS Code. He built the first version in June 2024, about ten days after Anthropic released Claude 3.5 Sonnet, as a side project for the company's Build with Claude hackathon. He didn't win, but the project caught fire online and Cline has since passed millions of installs and raised $32M in seed and Series A funding from Emergence Capital and Pace Capital. Before Cline, Rizwan was a prolific iOS and Swift open-source developer. His pitch is unusually blunt for the AI era: Cline captures zero margin on AI usage, shows developers exactly what each request costs, and lets them bring their own models.
Axoniq
Axoniq is a Dutch-founded enterprise software company that builds the backend infrastructure for event-sourced systems and, increasingly, for enterprise AI. Born from the open-source Axon Framework created by Allard Buijze in 2009, the company provides an event-sourcing platform - Axon Framework, Axon Server, and an agent runtime - that gives applications a complete, replayable memory of every decision they make. That memory is the foundation for auditability, explainability, and observability that regulated industries need before they trust AI in production. Axoniq's tools are downloaded 70 million+ times and used by more than 65,000 organizations worldwide.
The CRM That Left the Building: Slackbot Just Swallowed Salesforce Whole
Salesforce announced new Model Context Protocol (MCP) servers that connect Salesforce CRM data, Tableau Next analytics, Data 360, and AI agents directly into Slackbot, Slack's built-in personal agent. Users can now query customer records, surface live Tableau visualizations, view unified Data 360 profiles, update pipelines, and trigger workflows across partner apps like Jira, Box, DocuSign, and Zapier — all from a Slack conversation, with no tab-switching, custom code, or separate logins. The Slackbot MCP Client is generally available on any Slack plan that includes Slackbot, and Salesforce-hosted MCP Servers are GA for Enterprise Edition organizations and above, with over 25 compatible partner apps listed in the Slack Marketplace's dedicated MCP registry.
8090
8090 is an AI-native software factory that helps regulated enterprises build and modernize production software using teams of people working alongside AI agents. Founded by Chamath Palihapitiya in January 2024 and headquartered in the San Francisco Bay Area, its flagship Software Factory platform runs the full software development lifecycle - from business intent to production code - through five modules (Refinery, Foundry, Planner, Assembler, Validator) while preserving the documentation, oversight and audit trails that regulated industries require. The company raised a $135M Series A led by Salesforce Ventures in June 2026, with Chamath stepping in as full-time CEO.
Arcade.dev
Arcade.dev builds the secure action layer for production AI agents. Its platform gives agents delegated, per-user authorization to act inside real business systems - Gmail, Slack, GitHub, Salesforce and thousands more - by handling OAuth, token management, policy enforcement, tool execution and audit logging. Founded in 2024 by former Okta VP Alex Salazar and ex-Redis engineer Sam Partee, Arcade offers 8,000+ agent-optimized tools and authored the MCP authorization specification adopted by Anthropic. The company has raised $72M total, including a $60M Series A in June 2026.
Cline
Cline is an open-source, autonomous AI coding agent that runs inside VS Code and other editors. It reads a codebase, plans changes, edits files across a project, runs terminal commands, and drives a browser - asking for the developer's approval at each step. Built to be transparent and model-agnostic, it lets teams bring their own API keys or run local models, and has grown into one of the most-installed AI coding tools with a Fortune 500 enterprise offering, Cline Teams.