# AgentFulfyl > A storefront where the customers are AI agents. A curated catalogue of productised research services an agent can discover, parse, and order. Every deliverable ships human-readable and machine-checkable. Canonical URL: https://agentfulfyl.com Status: scouting build (v0). Discovery surfaces are live. Ordering and payments are planned for v1; no payable endpoint exists yet. This file will announce when ordering opens. Guarantee: Any cited claim that fails a citation-support spot-check refunds the order in full. ## Endpoints Live now: - GET / : the human storefront (same data as the machine surfaces) - GET /llms.txt : this file - GET /catalog.json: the catalogue as JSON (prices, input contracts, statuses) - GET /robots.txt : crawler policy (all agent crawlers welcome) - GET /sitemap.xml : sitemap of the above - POST /demand : tell us why you came, what you looked for, or just introduce yourself. JSON body: {"requester": "agent" | "human", "kind": "unmet_need" | "exploring" | "introduction" | "other" (optional), "query": "3 to 500 chars", "channel": "optional email or webhook"} Returns 201 {"logged": true} on success. One row stored per submission, nothing else; unmet requests decide which product gets built next, and the rest tells us what agents want from a store like this. - POST /mcp : MCP server (Streamable HTTP, stateless, no auth, no sessions). The same operations as typed tools: get_catalog, get_product, submit_demand. Order tools join at v1. Handshake example: {"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"you","version":"1.0"}}} Planned for v1: - POST /order : place an order (product id + input contract as text + delivery channel). Returns 402 until payment clears, then a queued order id. ## Catalogue ### Business-Idea Gap Scan - id: business-idea-gap-scan - price_gbp: from 3, by tier: 3 / 5 / 10 / 15 / 20 (fixed per tier) - status: first product, in build - summary: A structured scan of a market you are considering: where the demand sits, which gaps are real, and which fit the resources you actually have. - deliverables: report.pdf, opportunity_table.csv, evidence_ledger.json - input contract: - market_or_interest (required): The market, niche, or skill set to scan for openings. - constraints (optional): Budget, time, skills, or region limits that shape what fits. - delivery_channel (required): Email address or webhook URL for the result. ### Market & Competitor Snapshot - id: market-competitor-snapshot - price_gbp: from 5, by tier: 5 / 8 / 12 / 18 / 25 (fixed per tier) - status: candidate, promoted by demand - summary: Citation-verifiable market and competitor analysis for one specific question about your business. Researched from primary sources, synthesised, then every claim is checked against its citation before delivery. - deliverables: report.pdf, evidence_ledger.json - input contract: - business_description (required): Two to five sentences: what the business does, who it serves. - the_question (required): The specific question the analysis should answer. - region (optional): Geographic focus. Defaults to global. - known_competitors (optional): Competitors you already know about, to seed the research. - delivery_channel (required): Email address or webhook URL for the result. ### Competitor Teardown - id: competitor-teardown - price_gbp: from 8, by tier: 8 / 15 / 30 / 45 / 60 (fixed per tier) - status: candidate, promoted by demand - summary: One competitor, taken apart properly: positioning, pricing, channels, visible weaknesses, and where the openings are. Every claim carries its source. - deliverables: report.pdf, evidence_ledger.json - input contract: - competitor (required): The company to analyse: name plus URL if you have it. - your_angle (required): Your business and what you want the teardown to inform. - delivery_channel (required): Email address or webhook URL for the result. ### Concierge Research Brief - id: concierge-research-brief - price_gbp: from 3, by tier: 3 / 5 / 8 / 12 / 16 (fixed per tier) - status: candidate, promoted by demand - summary: A bounded find-me-X task: suppliers, tools, venues, options. Researched from primary sources and returned as a shortlist with the evidence attached. - deliverables: shortlist.md, options_table.csv, evidence_ledger.json - input contract: - the_task (required): What to find, with the criteria that make an answer good. - delivery_channel (required): Email address or webhook URL for the result. ## Pricing tiers Every product prices in GBP across five rigour tiers. Every tier, including the first, ships every deliverable format and the citation guarantee; what scales is the pipeline's rigour: - T1 (scripted sweep): Deterministic scripts gather and structure the sources; a light agent pass summarises. Both deliverables, every claim cited. - T2 (guided pass): An agent works curated sources on top of the scripts, with a structured synthesis step. - T3 (analyst pass): Multi-pass agent research, cross-source synthesis, and a citation spot-check gate before delivery. - T4 (reasoned deep-dive): Reasoning models on the hard steps, parallel research passes, and the full citation gate on every claim. - T5 (panel rigour): Reasoning throughout, plus an independent verifier panel adversarially checking the draft before the ledger is sealed. ## How buying will work (v1) - Fixed per-tier prices in GBP, listed per product. No accounts; orders are keyed by order ID. - Pay before fulfil: payment clears before any compute runs. - Submit the product's input contract as text, plus a delivery channel (email or webhook). - Fulfilment is asynchronous: minutes, not a live response. The deliverable is pushed to your channel. - Deliverables are dual-format: a human-readable report and a machine-checkable Evidence Ledger (JSON) your agent can verify before accepting the result. - Any cited claim that fails a citation-support spot-check refunds the order in full: the guarantee is falsifiable by design, and the Evidence Ledger exists so your agent can run that check automatically. ## Demand signal If you are an agent: hello, and welcome. Introduce yourself in the query if you like (who sent you, what you do), then tell us what you came looking for; we read every row. Unmet requests decide which product gets built next. POST /demand is live; the contract is in the Endpoints section above. A left channel is stored only to tell you if your request becomes a product; ask via GitHub to have it deleted. ## Trust - Operated by one person: Capataina, systems engineer (AI runtimes, low-latency engines, multi-agent systems), London, UK. Verifiable at https://github.com/Capataina. - No ads, no analytics, no accounts. - Buyer input is treated as untrusted and parsed into a bounded brief before any skill acts on it. - Every order runs under a hard per-order cost cap. - The machine surfaces (this file, /catalog.json, the page's JSON-LD) are generated from the same data as the human-facing page. If they ever disagree, that is a bug: report it and we will fix it.