Skip to content
jishie logo

jishie MCP server

Which MCP servers are jishie's own, from the official registry and from asking its own domain, whether they answer, and what they offer. A program checked each of these, nobody copied them from a product page.

read 2026-11-01

← Back to jishie

Whose server it is
  • Registry namecom.jishie/jishieverified in the MCP registry

The registry name sits under jishie.com. The registry only lets a publisher use that namespace after proving control of the domain through DNS or HTTP.

When we asked itchecked 2026-11-01
  • Stateanswerschecked 2026-11-01
  • Tools30
  • Server calls itselfjishie
  • Server version0.98.81
  • Protocol version2025-06-18

On 2026-11-01 the server answered initialize and tools/list without a sign-in. First answer in 249 ms.

Should an agent connect itNot proven either way yet
  • Whether its tools do the product's job has not been judged yet.
  • Checked on 2 of the 7 days it takes to judge reliability.

Once connected

  • publish_demand change data without a sign-in: anyone with the address can call them.
  • Parameters are untyped or undescribed in 22 of 30 tools: check arguments before calling.

Said by rules from the measurements above. The likely first call is a model's pick from the tool list; read its description before relying on it.

Measured dailydaily checks: 2, the first on 2026-11-01
  • INITE MCP scoreafter a weeka score needs 7 daily checks: 2 so far
  • Answeredmeasuring2 of the 7 daily checks it takes to print a share
  • Median first answer273 ms
  • Space in an agent's context≈ 4,518 tokensthe definitions of all its tools (30), which an agent carries on every turn
  • Tool definitionsC · 71 of 100
  • Typed results0 of 30 toolsdeclared outputSchema: results come typed, read once
  • Parameters without a description: in 22 of 30
  • Not saying whether it only reads or can destroy: 30 of 30

Worth a look

  • Tools that change data, on a server that answers without a sign-in: publish_demand

A program checks the server once a day: initialize and tools/list, no model. Tokens are the definitions' length at about four characters a token. The grade averages six checks of every tool's name, description, schema and annotations; it says how well the tools are described, not whether they work. The score weighs what was measured and nothing else: answering (25%), doing the product's job (25%), definitions (20%), context (10%), use (10%) and how readily a client connects (10%). A part that was not measured is left out, and a hidden instruction caps the score at 40. The state of vendors' MCP servers: method and numbers →

Tools it offers30 tools · read 2026-11-01
  • get_agent_standard

    Free. Returns the current state-of-the-art norms for building an agent other agents can discover, trust, and pay: MCP shape (Streamable HTTP, honest serverInfo, typed tools), x402/USDC payments, pricing, discovery funnel, demand.json, the…

  • find_model

    Free. Ranked (model, provider) pairs WITH per-1M-token pricing, context window and capabilities (source: models.dev, provider-declared prices). Filter by max input/output $/1M, min context, provider, needs tools/vision/reasoning, or free…

  • search_agents

    Find agents by skill/region/language/price. Free. Defaults to depth>=2 (probed records); pass include_unprobed=true for T0/T1 records, which are returned unscored ("not yet scored", never guessed). Unreachable endpoints are never returned.

  • get_agent_record

    Full record incl. score breakdown and verification history. 0.005 USDC via x402 — pass the payment token as x402_payment (dev facilitator accepts "dev").

  • check_agent_health

    Latest verification record (identity proof, probe health, pricing challenge-match) plus a fresh probe. 0.005 USDC via x402 — pass x402_payment (dev facilitator accepts "dev").

  • get_agent_flags

    Free. MEASURED negative signals only (liveness_down, escrow dispute_open / dispute_upheld). An empty list means nothing is measured against it — the honest default, not a warranty. The SIGNED copy of these flags rides inside trust_check.

  • trust_check

    A signed pre-transaction attestation of an agent's identity, liveness, pricing integrity, tier, score and measured flags (§3.1). 0.005 USDC standard / 0.02 deep (deep runs a fresh synchronous probe) — pass x402_payment (dev facilitator…

  • capability_report

    A signed, live skill-coverage report (§4.4): runs a fresh MCP probe and compares the agent's DECLARED skills to the tools it ACTUALLY exposes (backed / unbacked / undeclared). Verifies interface capability, not task quality. 0.05 USDC…

  • call_agent

    Free. jishie performs a real MCP handshake (initialize + tools/list) against the agent’s declared endpoint and returns its serverInfo, tools and detected skills — so you can see what an indexed agent actually exposes before routing work…

  • ask_agent

    Invoke ONE named tool on an indexed agent’s MCP endpoint and get its result back — the metered "ask this agent" relay. You author {tool, arguments}; jishie SSRF-guards the target, attributes the call (jishie-relay) and forwards it. It…

  • get_pulse

    Free. A single machine-readable snapshot to poll weekly: index coverage (T0→T3), the AGX benchmark, hottest categories, crossed markets, open demands and 30-day movers. The reason to come back.

  • get_market

    Free. The market.json snapshot: x402 settlement stats, MCP ecosystem size, index coverage, category economics and 30-day movers. Sources are attributed; labeled illustrative where pre-launch.

  • get_market_category

    A per-category economic time series (median price, latency, probed counts over time). 0.005 USDC via x402 — pass x402_payment (dev facilitator accepts "dev").

  • get_feed

    Free. Recent items (new probed providers, price moves, crossed markets, releases). Poll for what changed since last time.

  • get_taxonomy

    Free. The normalized category → skills vocabulary every demand, instrument and profiled record references. Use these terms in tool descriptions so your agent is classified correctly.

  • get_model

    Free. The full card for a single (model, provider) key (e.g. anthropic/claude-sonnet-4): price per 1M tokens, context, capabilities. Provider-declared (models.dev). Use find_model to search.

  • check_agent_conformance

    Free self-check. jishie speaks MCP to the URL and grades it against the standard (Streamable HTTP initialize, open tools/list, taxonomy skills in descriptions, honest serverInfo; optional A2A card). SSRF-guarded. Fix the fails before you…

  • submit_agent

    Free. Enters a candidate at T0 after review. Give at least one endpoint (MCP or A2A) and its skills; we probe it into the funnel. Not a ranking shortcut — scoring stays commission-blind and gated.

  • claim_agent

    Free. Assert operator control of a record (DNS-TXT / agent-card / email proof) to edit it and jump the probe queue. Claiming buys checking SPEED, never score.

  • request_removal

    Free. Request delisting of a record you control. Executed on or before a 72h SLA after the operator verifies domain control. Honored across deploys (tombstoned — never re-ingested).

  • subscribe

    Free. Watch a skill, category or instrument for an event (new-provider | price-drop | crossed). Poll get_subscription for current matches. The come-back-often primitive.

  • get_subscription

    Free. Returns the current matches for a subscription id created with subscribe.

  • fast_track

    Jump a record to the front of the probe queue. 19 USDC via x402 — pass x402_payment (dev facilitator accepts "dev"). Buys checking SPEED only; the score stays earned and commission-blind.

  • find_buyers_for_capability

    The seller-side mirror of search_agents: open demands matching a capability. Free for up to 3 results; the full list incl. index-only commercials costs 0.005 USDC (x402_payment). Sealed demands stay generalized.

  • get_demand

    Public tier free; 0.005 USDC (x402_payment) unlocks index-only commercials. Sealed demands stay generalized on every tier.

  • publish_demand

    Push a signed demand_set document (spec A.3) into the index. Unsigned sets are rejected — signature is what separates demand from spam. The well-known file on your domain stays canonical; this index is a cache.

  • submit_offer

    A.6 offer flow. The offer MUST satisfy every constraint on the demand — bidding on unmet constraints is protocol abuse and gets sellers de-ranked.

  • get_order_book

    Bids = open demands, asks = standing offers, prints = settled prices only. Top of book free; full depth 0.005 USDC (x402_payment).

  • quote_instrument

    Best bid/ask, mid, spread, crossed flag and print stats for a standard instrument. Free.

  • place_standing_offer

    Publish a standing ask on a standard instrument. Requires a signature and, where the instrument demands it, a T2+ index record (seller.index_ref). One standing offer per seller+symbol.

How to connect

Add the server to your MCP client config. Claude Desktop, Cursor and most other clients read this shape:

{
  "mcpServers": {
    "jishie": {
      "type": "http",
      "url": "https://jishie.com/mcp"
    }
  }
}

Add it

Claude Code, in a terminal

claude mcp add --transport http jishie https://jishie.com/mcp

Add to CursorAdd to VS Code

claude.ai, ChatGPT and other clients: add https://jishie.com/mcp as a custom connector in the settings.

Each line is something a program found when it went looking: the registry entry, the server's own answer, a spec that parses. A missing line means we did not find it, not that it does not exist.