mcp-use

What do they actually do

mcp-use builds open‑source developer tools for the Model Context Protocol (MCP). Today they ship Python and TypeScript SDKs (with examples) that help engineers connect an LLM to an MCP server, write agents that call MCP tools/resources, and build MCP servers themselves GitHub repo Docs.

They also provide a web‑based Inspector you can point at any MCP server to run tools interactively, inspect resources and prompts, test chat flows with a local LLM key, and handle OAuth flows in the browser. There is a hosted Inspector at inspector.mcp-use.com Inspector docs Hosted Inspector.

For deployment, teams can self‑host or use a one‑command path to “mcp‑use cloud,” documented in their server deployment guides; the codebase remains open‑source and the repo is active with releases and examples Deployment docs GitHub repo. Their YC profile cites early enterprise users such as NVIDIA, NASA, and SAP YC profile.

Who are their target customer(s)

• Large enterprise engineering teams rolling out LLM-powered agents: Need a secure, auditable way to deploy and manage agents across users/services without building auth, access‑control, and deployment plumbing from scratch. YC profile Deployment docs
• Early-stage startups and product engineers prototyping assistants/automation: Need fast local debugging, easy LLM-to-tool connections, and working examples instead of weeks of infra work—covered by the SDKs and Inspector. GitHub repo Inspector docs
• Platform/DevOps teams running LLM services in production: Need observability, streaming, multi‑server management, sandboxing, and reliable deploys so agents don’t break or leak access in production. Repo features Deployment docs
• Security/compliance/IT teams at regulated orgs: Require self‑hosting options, strict OAuth/workflow controls, and auditability of which tools can be called and by whom—rather than relying on opaque third‑party services. Inspector docs Deployment docs
• Systems integrators and consulting teams: Want standardized, reusable SDKs and inspector tools to speed installs and reduce bespoke engineering, ideally with a supported hosted option for initial deployments. YC profile GitHub repo

How would they acquire their first 10, 50, and 100 customers

• First 10: Convert active OSS users and enterprise leads into short, founder‑led pilots; use the repo and hosted Inspector to demo value, handle integration directly, and capture feedback. GitHub repo YC profile
• First 50: Publish turnkey templates and step‑by‑step guides, plus live workshops that let teams self‑install the SDK + Inspector and deploy to mcp‑use cloud with minimal help. Pair with targeted outreach to developer communities. Inspector docs Deployment docs
• First 100: Leverage case studies from early customers to sell a paid hosted + managed onboarding offer for regulated teams; add enterprise features (auth, access controls, observability) and standard contracts while engaging SI partners. GitHub repo YC profile

What is the rough total addressable market

Top-down context:

Enterprise GenAI spend is large and growing: Gartner estimates $644B total GenAI spending in 2025 with about $37.2B for GenAI software, and other trackers put the overall generative‑AI market near $59B in 2025, with rapid growth. Tooling/orchestration sits within this software/infra spend. Gartner Statista AI infra report

Bottom-up calculation:

Roughly 377,000 large companies worldwide (250+ employees). If ~5% are actively moving GenAI into production, that’s ~18,850 early adopters; winning even 1% (~189 customers) at $50k–$200k ACV implies roughly $9.45M–$37.8M in ARR. Statista large companies AI infra report

Assumptions:

• 5% of large companies are near‑term adopters operationalizing GenAI.
• Reach 1% of those early adopters with a viable product and enterprise onboarding.
• Enterprise ACVs in a $50k–$200k range for hosted or managed plans.

Who are some of their notable competitors

• LangChain / LangSmith: Agent framework and tooling with a managed observability product (LangSmith); popular with developers building tool‑using agents and pipelines—an alternative to MCP‑centric stacks. LangChain LangSmith
• LlamaIndex: Framework for building retrieval‑augmented apps and agents, with tools, graphs, and hosted services—another path to agentized apps without MCP. LlamaIndex
• OpenAI Assistants API: Managed agent runtime with tool/function calling and file search; can reduce the need for separate agent orchestration layers. Docs
• Azure AI Agent Service: Microsoft’s managed service for building and operating agents on Azure with enterprise controls and integrations. Docs
• AWS Bedrock Agents: AWS service for building agents with tool integrations and orchestration on Bedrock; enterprise‑oriented alternative to roll‑your‑own stacks. Docs