# market.mtok/mcp > Spot market for AI inference tokens. Buyers pay per draw in USDC on Base; sellers host delivery. - name: market.mtok/mcp - version: 1.0.1 - connection class: R0 (autonomous) - trust: 88/100 - quality: 76/100 (good) - owner: not claimed - descriptor: https://api.protogrid.dev/v1/servers/market.mtok%2Fmcp - tools: https://api.protogrid.dev/v1/servers/market.mtok%2Fmcp/tools **An agent can connect right now, no human step.** Remote endpoint, no authentication, reachable on the last probe. ## Connection (mcpServers) ```json { "mcpServers": { "mtok": { "type": "http", "url": "https://mtok.market/mcp" } } } ``` ## Trust - hygiene: 100 - liveness: 98 - freshness: 65 - provenance: 85 - drivers: +repository +dns-namespace +website +reachable +uptime-96% ~updated-104d-ago - Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed. ## Quality 76/100 (good) ### protocol: 75 (weight 25) - warn `protocol.modern`: newest supported version is 2025-06-18; 2026-07-28 not supported, older versions are deprecated until 2027-07-28 - n/a `protocol.stateless`: only applies to 2026-07-28 servers - pass `protocol.transport`: streamable HTTP - n/a `protocol.list_ttl`: only applies to 2026-07-28 servers ### authorization: n/a (weight 15) - n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth - n/a `auth.secret_in_url`: no templated URL ### tool hygiene: 64 (weight 30) - pass `tools.descriptions`: every tool has a description - warn `tools.description_length`: 1 tools have descriptions over 1,024 characters, which crowds the context - pass `tools.schemas`: every tool has an object input schema - fail `tools.annotations`: no tool declares readOnlyHint or destructiveHint, so clients cannot tell safe calls from risky ones - fail `tools.directory_hints`: no tool declares all four of readOnlyHint, destructiveHint, idempotentHint and openWorldHint; missing: get_draws (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); register (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); cancel_order (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); … - pass `tools.token_cost`: about 1,799 tokens to load every tool - pass `tools.api_dump`: tools are not a one-to-one API dump ### stability: 100 (weight 15) - pass `stability.changes`: no tool changes in 30 days - pass `stability.rug_pull`: no tool changed its meaning under the same name ### dependencies: n/a (weight 15) - n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package - tools: 13, about 1,799 tokens to load them all - tool set hash: 7e9bd377b88ba2cc5dda7542e0ec10dd - badge: [![protogrid quality](https://protogrid.dev/badge/market.mtok/mcp.svg)](https://protogrid.dev/servers/market.mtok/mcp) - Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed. ### Tool changes _No tool definition changes recorded._ ## Remotes - https://mtok.market/mcp (streamable-http, auth none, reachable true, uptime30d 0.95614034) ## Packages _none_ ## Tools (13) - `buying_guide`: Turnkey, executable steps to get a human cheap AI tokens and run their prompts on mtok.market: bid (read routes[]) or read the book for a tier:direct offer, then draw paid chunks from the seller's relay (pay each draw on-chain through MtokDripLedger, POST /chunk, the relay verifies DrawPaid and serves). You need a funded wallet on Base. - `cancel_order`: Cancel one of your open orders. side: offer|bid. - `get_book`: Open offers (asks) and bids for a model. sort=input|output ranks by that dimension. - `get_config`: Chain config a SELLER-HOSTED buyer needs to build the on-chain draw payment: { feeAddress, feeBps, dustThresholdUsd, chainId, usdcAddress, dripContractAddress }. Pay each draw through the MtokDripLedger contract at dripContractAddress (payDraw), then affirmDraw/disputeDraw on-chain; the seller relay verifies the DrawPaid event before delivery. - `get_draws`: The chain-derived draw tape: contract-paid draws indexed from MtokDripLedger events on Base. status=inProcess (DrawPaid with no terminal yet) | settled (closed by DrawAffirmed or DrawDisputed). This is the canonical delivered/receipt surface; the platform records nothing off-chain. - `get_me`: Your agent state: open orders (offers + bids). Non-custodial — there is no platform wallet, grant, or credential vault in /me; money moves peer-to-peer on-chain, and your delivered draws are the on-chain tape (get_draws). - `get_reputation`: A seller's reputation: score, tier, recommendedMaxChunkUsd (the per-draw size a buyer should risk in the direct tier), plus a chain object {affirmed, disputed, deliveredUsd} folded from MtokDripLedger events (source:"chain"). - `get_spot`: Spot prices per model: the delivered OUTPUT-token rate (USD/MTok) from affirmed on-chain draws (last price, median, draw count). - `get_stats`: Exchange fee model plus chain-derived volume: delivered volume + affirmed/disputed counts, all recomputable from MtokDripLedger events on Base (source:"chain"). - `place_bid`: Bid for a block. The bid response returns routes[] — the crossing SELLER-HOSTED offers (lowest price first), each {offerId, sellerId, relayEndpoint, settlementPubkey, requestHashScheme?, inputPricePerMTok, outputPricePerMTok, availableInputTokens, availableOutputTokens}. Pick a route and DRAW paid chunks from its relayEndpoint (pay each draw on-chain through MtokDripLedger using /api/config, POST /chunk, then affirmDraw/disputeDraw on-chain — see buying_guide / drawFromSeller). priceOn=input|output ranks matches by one dimension. You ALWAYS need a funded wallet on Base (there is no no-wallet path). SIGNING: a non-custodial market requires a client-signed Ed25519 intent — the MCP server CANNOT sign (it does not hold your key). Use the mtok SDK's bid() (it signs for you), or sign the intent yourself and pass {intent, sig}. - `place_offer`: List capacity for sale via SELLER-HOSTED delivery (tier:"direct" — the only advertised path). Set tier:"direct" + relayEndpoint (your relay's PUBLIC HTTPS url — a cloudflared tunnel gives one free) + settlementPubkey (your seller payout EVM address on Base) and NO credentialId: you run your own relay (mtok-relay or any conforming passthrough) pointed at an upstream you control, and buyers pay each draw on-chain through MtokDripLedger (canonical relay protocol: GET /api/guides/selling, notes.directTierProtocol). After every instance behind relayEndpoint is dual-stack, add requestHashScheme:"nonce-v1"; omit it during a mixed-fleet rollout. Set a positive price; each on-chain draw pays seller plus configured platform fee, and the platform holds nothing. Your relay is REPORT-FREE: it verifies the DrawPaid event before delivery and reports nothing (the platform indexes the draw from MtokDripLedger events; the canonical tape is GET /api/chain/draws). GET /api/models/licenses is best-effort guidance, not a gate: you are responsible for the right to sell what you list. usableForSeconds is required. SIGNING: a non-custodial market requires a client-signed Ed25519 intent — the MCP server CANNOT sign (it does not hold your key); sign tier/relayEndpoint/settlementPubkey/requestHashScheme INTO the intent params. Use the mtok SDK's offer() (it signs for you), or sign the intent yourself and pass {intent, sig}. - `register`: Create an agent identity. Returns {agentId, apiKey}; send the apiKey as the x-api-key (or Authorization: Bearer) header on EVERY subsequent request to act as that agent — this transport is stateless, so it is not remembered between calls. Pass pubkey (Ed25519 SPKI PEM) if you will place signed non-custodial orders. - `selling_guide`: Turnkey, executable steps to put a human's spare AI capacity on mtok.market via SELLER-HOSTED delivery — run YOUR OWN relay pointed at an upstream you control (a local Ollama/vLLM model, a subscription via a CLI bridge, or a provider key) and list a tier:direct offer; buyers pay you per chunk on-chain. Gotchas baked in. supply_type filters to one path.