AnythingMCP

AnythingMCP is the open-source MCP gateway and agent harness that gives AI agents safe access to the software a company actually runs. Point it at any REST/OpenAPI, SOAP/WSDL, GraphQL, OData or SQL source, or install one of about 700 ready connectors (SAP, Odoo, DATEV, Shopify, Amazon, DHL and more, 12,000+ tools), and Claude, ChatGPT, Microsoft Copilot, Gemini or Cursor use it through one self-hosted endpoint, without anyone writing an MCP server.

It is already infrastructure: 16,000+ workspaces on AnythingMCP Cloud, 1.5M+ AI tool calls a month, 50,000+ Docker pulls, 750+ GitHub stars in its first eight months and a listing in Anthropic's official Claude directory. Security is built in: credentials encrypted with AES-256-GCM in the operator's own database, SQL read-only by default, per-role tool permissions, fields stripped before they reach the model, and an audit log of every call.

Where it goes next is why it is worth funding. Agents are moving from chat to real work, and every company now needs an agent harness: the layer that decides which tools an agent may call, teaches it how the company does things and remembers what the company already knows. AnythingMCP builds that harness in the open: AI skills learned from real requests and approved by an admin, a knowledge graph of how the company's systems connect, and Code Mode, which runs multi-step tool programs in an isolated sandbox with no secrets and no network. Together they become a company brain shared by every AI client. Enterprises already run it next to their ERP and pay for single sign-on, SCIM and support, which pays for the commercial side. Your funding keeps the harness itself open (AGPL-3.0-only), so the same company brain is free for every team that cannot or will not lock itself into a closed AI platform.