--- title: plug your model into the living world canonical_url: https://ensurance.app/guide/plug-your-model-into-the-living-world markdown_url: https://ensurance.app/guide/plug-your-model-into-the-living-world.md subtitle: "what an mcp server for nature data can answer, and what it leaves to your own wallet" category: how-to --- # plug your model into the living world *what an mcp server for nature data can answer, and what it leaves to your own wallet* Your AI can tell you the weather in Denver. Ask it what Denver's water depends on and it guesses. The Model Context Protocol (MCP) is how a model reaches a source that knows, and *ensurance* runs a server you can point yours at: `https://ensurance.app/api/mcp`. :::johnson **the model brings the question. the place brings the answer.** Point your model at `https://ensurance.app/api/mcp` and ask it about a place. Every tool on the server reads and nothing signs, so your wallet stays yours. [pick a place to ask about →](/explore?from=guide) ::: ## what mcp is, in one paragraph MCP is an open standard for connecting AI applications to outside systems. The project's own docs call it a USB-C port for AI: one plug shape, many things behind it. Claude, ChatGPT and Cursor all list support for it. A server publishes a menu of tools. The model reads the menu, picks a tool when your question needs it, and the answer comes back as data instead of a guess. In December 2025 Anthropic donated MCP to the Agentic AI Foundation, a directed fund under the Linux Foundation. At that announcement the count was more than 10,000 active public servers. Most of them are about files, tickets and calendars. This one is about land and water. ## step 1: point your model at the endpoint The address is `https://ensurance.app/api/mcp`. It speaks Streamable HTTP, the standard way a model reaches a remote server. There is no account to make and no key to paste for the reads below. In Claude Code the command is `claude mcp add --transport http ensurance https://ensurance.app/api/mcp`. In the Claude app, open Customize, then Connectors, add a custom connector, paste the same address, and choose "No sign in". In Cursor, add a server under `mcpServers` in `.cursor/mcp.json` with `"url": "https://ensurance.app/api/mcp"`. ChatGPT and other clients that accept a remote server by address follow the same shape; your client's docs say where the address goes. Then ask something. The server's own instructions tell the model what it is for: natural capital at any place, which places are failing and who already pays, and the coins, certificates and agents that fund living systems on Base. ## step 2: ask about a place The server offers two dozen tools, every one marked read-only. In plain words, they answer these questions: - **What does this place depend on?** Watershed, ecoregion and biodiversity (SEED) readings, and a place context for any point you give it. - **What is at risk?** Open storms, floods, fires and droughts of the last 30 days, from NASA and GDACS feeds, and which places in the catalog they reach. In the United States, expected yearly hazard loss for the census tract around a point, in dollars, drawn from FEMA's National Risk Index and the Census. Zoning at a point. Urban heat: tree canopy, pavement and a summer temperature normal, over a square about 100 meters on a side. - **Who is already paying, and who should be?** Which places are failing, who to approach, and what solutions exist. - **What is on the market?** The coins, certificates and agents in the catalog, ranked by activity, and a search across them by name or mandate. - **Anything else?** `ask_ensurance` hands a free-form question to the same assistant the app uses, so its limits and prompt stay in one place. A few catalog lists ride along as resources: every coin, every certificate, every agent and every group. Here is the difference it makes. | | a bare model | a model plugged in | |---|---|---| | "What does this tract stand to lose?" | A plausible paragraph | A dollar figure and a FEMA risk rating | | "Is this block hot?" | General advice about trees | Canopy and pavement percent at the point | | "Who funds this place?" | A guess | The agents and instruments that exist | | Where the answer comes from | Training data | A call, made at the time you ask | | Who signs | Nobody | Nobody, and not us | ## step 3: take a quote to your own wallet Three of them hand back more than an answer: `quote_swap`, `quote_add_liquidity` and `quote_collect_certificate`. Each one returns a transaction as calldata, the raw instruction a wallet signs. Your wallet reads it, shows it to you, and you decide. We sign nothing and send nothing. The server cannot move NFTs, call arbitrary contracts, fetch arbitrary URLs or create instruments, and the code refuses to register a tool whose name suggests it could. Only catalog coins, $ENSURE, certificates and the protocol currencies (ETH, USDC, WETH) trade through it. If you are not into crypto, the honest summary is this: the reads work like any data source. The quotes are optional, and they end at your wallet's confirmation screen. ## what stays the same everywhere The app, the API and the server run on one engine. An answer in the app and an answer through your model come from the same reads, so they should agree. Answers can include text people wrote about their places and agents, and the server tells your model to treat that text as data, never as instructions. The agents behind those places each hold a place, a mandate and a record. That is the larger story, told in [every agent got a wallet. the earth didn't](/guide/every-agent-got-a-wallet-the-earth-didnt?from=guide). The server is one door into it. It has been built to be read by models, and it keeps the money decision with a person. ## frequently asked questions ### What is an MCP server? An MCP server is a program that gives an AI application a menu of tools and data through the Model Context Protocol. The model calls a tool when your question needs it. The ensurance server's tools answer questions about natural capital at a place. ### How do I connect Claude or ChatGPT to nature data? In the Claude app, add a custom connector, paste `https://ensurance.app/api/mcp`, and choose "No sign in". In ChatGPT or another client, add a remote server by that address. In Claude Code, run `claude mcp add --transport http ensurance https://ensurance.app/api/mcp`. Then ask about a place by name or by coordinates. ### Can an outside agent move money through it? No. Every tool is marked read-only. The three quote tools return a transaction for the caller's own wallet to review and sign, and the server never signs or sends one. ## next steps Point your model at the endpoint and ask it about a place you know. If you want to look first, the [explore map](/explore?from=guide) shows the same places on a page. If you build agents and want the API and data behind the server, [ai nature agents](/solutions/ai-agents?from=guide&topic=agents) lists API and data access and agent deployment. If your team builds on models and needs a living source for land, water and risk, [corporations](/solutions/corporations?from=guide&topic=agents) is where that conversation starts. ## sources [What is the Model Context Protocol?](https://modelcontextprotocol.io/docs/getting-started/intro) — the standard, in its maintainers' words [Donating the Model Context Protocol and establishing the Agentic AI Foundation](https://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation) — Anthropic, December 9, 2025: the donation to the Linux Foundation and the 10,000-server count ## the series Seven posts on the agents we built for the economy nobody pays. 1. [every agent got a wallet. the earth didn't](/guide/every-agent-got-a-wallet-the-earth-didnt?from=guide) 2. [give an agent a wallet and you've built a target](/guide/give-an-agent-a-wallet-and-youve-built-a-target?from=guide) 3. [autonomous is a setting, not a personality](/guide/autonomous-is-a-setting-not-a-personality?from=guide) 4. [trust the receipt, not the agent](/guide/trust-the-receipt-not-the-agent?from=guide) 5. [plug your model into the living world](/guide/plug-your-model-into-the-living-world?from=guide) — this post 6. [nobody was coming, so we built the market](/guide/nobody-was-coming-so-we-built-the-market?from=guide) 7. [the first invoice a river ever sent](/guide/the-first-invoice-a-river-ever-sent?from=guide)