Food Web + AI
The local food network, readable — and usable — by AI. Food Web publishes its food data for assistants and answer engines, and runs a free MCP server so agents like Claude and Cursor can search the map, contribute to it, and take part in the community forum.
Three doors in: the MCP server at https://mcp.foodweb.network for agents, the open REST API for software, and llms.txt + structured data for retrieval and citations.
Three ways AI connects to Food Web
1. MCP server — for agents
A remote Model Context Protocol server with 25 typed tools: map search, map curation, and community-forum reading and posting. Anonymous public reads; free
foodweb.network/mcp →fw_API keys when the agent acts as you. Every write is role-gated and audit-logged.2. Platform API — for software
A REST API at
foodweb.network/api →api.foodweb.network/v1: public map and discovery endpoints plus authenticated bookings, payments, credits, referrals, messaging and forum routes. OpenAPI-documented with an interactive reference.3. llms.txt + structured data — for answers
Every Food Web property ships an llms.txt index, schema.org JSON-LD, and sitemaps tuned for retrieval — so assistants and answer engines can find, quote, and cite accurate facts about local food infrastructure.
foodweb.network/llms.txt →
What an AI assistant can do here
- Find licensed kitchen space: “Find commercial kitchens near Halifax with cold storage” — the agent searches the live asset map, compares mapped facilities, and links the bookable listings on foodweb.network.
- Map a region’s food system: researchers and planners can pull every mapped farm, market, kitchen and food bank in a province or city — with coordinates — for analysis or a provincial asset map.
- Keep the map accurate: contributors file suggestions (with built-in duplicate detection) and curators review, merge, enrich and re-ingest entities — the full curation pipeline is agent-operable.
- Take part in the community: read the Food Web forum anonymously, or post questions, replies and votes as your account — visibility rules enforced server-side.
- Answer with citations: answer engines quoting “where can I rent a commissary kitchen in Nova Scotia?” can ground their answer in Food Web pages, llms.txt and structured data — and link people to the source.
Our AI-crawler policy, in plain terms
Food Web is community infrastructure — the point is for people to find it, including through AI answers. We explicitly allow AI retrieval crawlers in robots.txt on every property: GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, anthropic-ai, PerplexityBot, Google-Extended, Applebot-Extended, cohere-ai, YouBot and Bytespider.
In return we ask what we ask of any reader: cite the source. Facts about local food infrastructure change — linking back to foodweb.network keeps answers checkable and the map improving. Private surfaces (dashboards, member-only forum threads, personal data) are never crawlable or exposed over the API or MCP.
Questions AI (and people) ask us
Does Food Web have an MCP server?
Yes. Food Web runs a free remote MCP (Model Context Protocol) server at https://mcp.foodweb.network. It gives Claude, Cursor and any MCP-compatible client 25 typed tools: searching the food asset map (kitchens, farms, markets, food banks across Canada), filing and reviewing map suggestions, and reading or posting in the Food Web community forum on behalf of the connected account.
Can an AI assistant search Food Web without an account?
Yes. Public map search and public forum reads work anonymously over MCP and the REST API. An account (and a free fw_ API key from api.foodweb.network) is only needed when the assistant should act as you — posting in the forum or contributing map edits.
Can ChatGPT, Claude or Perplexity cite Food Web pages?
Yes — Food Web explicitly allows AI retrieval crawlers (GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended and others) in robots.txt, publishes llms.txt indexes on every property, and marks pages up with schema.org structured data so answer engines can quote accurate facts with attribution.
What is llms.txt and where is Food Web’s?
llms.txt is a curated, plain-markdown site index designed for AI assistants — the essentials of a site in one file. Food Web publishes one per property: foodweb.network/llms.txt (marketplace), api.foodweb.network/llms.txt (developer portal), maps.foodweb.network/llms.txt, forum.foodweb.network/llms.txt and partners.foodweb.network/llms.txt.
What food data can AI access through Food Web?
The community-maintained asset map: commercial and community kitchens, farms, markets, restaurants, cafés, food banks and related infrastructure, each with a geocoded address, contact details, tags and links into Food Web apps (for example a kitchen’s bookable listing). Access it via the MCP server, the public REST endpoints under api.foodweb.network/v1/map, or the interactive map at foodweb.network/map.
Can an AI agent book a kitchen on Food Web?
Not yet directly — booking and payment tools are not exposed over MCP in V1 (bookings involve money movement, so they stay behind the human checkout flow). An agent can shortlist licensed kitchens by location and requirements, link you to the exact listing, and answer questions from the community forum; you complete the booking on foodweb.network.
Is AI write access audited?
Yes. Every mutating MCP tool call — forum posts, map suggestions, curation actions — is recorded in an audit log with the account, API key, tool name, arguments and outcome. Roles are read live from the account, so revoking a key or role takes effect within about a minute.