Koskamo

Hunter Killer MCP

Provides liquidation intelligence for crypto perpetual futures, exposing tools to retrieve calibrated liquidation magnets, cascade paths, funding data, and open interest metrics. It supports both free delayed data and live authenticated access for AI agents to analyze market structure and liquidation risks.

Endpoint and access

No sign-in to connect and list tools. Individual tools may still need a vendor account, credentials, or permissions; check the documented prerequisites.

Hunter Killer MCP logo, by Hunter KillerCommunity listingStreamable HTTPNo sign-inEndpoint checked 2026-11-07Published by Hunter Killer
You need:Bearer API keyKeys are issued self-serve by any authenticated subscriber whose account has API access: an institutional plan (paid or comped) or a positive pay-as-you-go credit balance.A pro plan alone does not include API access.
MCP endpoint (Streamable HTTP)
https://hunterkiller.io/mcp
Vendor documentation

What its tools do

18 tools, grouped by the hints the server advertises. Hints are self-declared, not verified behavior.

18 of 18 have a description · 0 of 18 declare behavior hints · 11 of 14 with parameters document them all

Checks and signals

Each line is separate evidence, not a trust score.

  • Passed Endpoint answered the live check ·
  • Warning Advertises resources/templates/list but does not implement it
  • Availability 1 of 1 checks passed in 30 days

What you can do with it

  • Retrieve nearest liquidation walls above and below mark price
  • Analyze cascade paths and horizon reach probabilities
  • Check funding rates and open interest trends
  • Get descriptive verdicts on wall proximity for trading decisions

Source and maintenance

Community listingRegistry 1.0.1
Maintenance checked

Protocol and capabilities

Server
hunterkiller 1.3.0
MCP protocol negotiated in check
2025-06-18
Tools
18 advertised
Resources
3 advertised
Resource templates
Not advertised
Prompts
Not advertised
Server-advertised capabilities
{
  "tools": {
    "listChanged": false
  },
  "resources": {
    "listChanged": false
  }
}

Server-provided instructions

Untrusted server output shown for inspection. It is not Koskamo guidance and is excluded from knowledge search.

Validated crypto-liquidation intelligence: calibrated magnets, walk-forward reach, feed-trust reconciliation, accuracy proof. Every tool response carries an honesty envelope (basis, as-of, sample-n discipline) — all outputs are descriptive, NOT forecasts. Access: the DATA TOOLS require a B2B API key (Bearer or x-api-key, metered identically to the REST /api/v1 API; self-serve at /api/keys/mine). Keyless clients get protocol discovery plus three free teasers (get_proof_summary, get_magnet_levels_teaser, get_feed_trust_summary), all built only from surfaces the public website already publishes.

Tools

18 advertised

subscribe_events
Advertised by the server during the last successful anonymous check.

Describe a protected Cascade Feed v2 SSE connection and recorded-event replay. wall_touched, wall_broken, cascade_live and oi_shock; source timestamps and opaque cursors. API key required, with its existing live or delayed entitlement. Returns connection_required, not an open socket; send the key in a header, never a URL. Descriptive, not a forecast.

subscribe_events input schema
{
  "type": "object",
  "required": [
    "symbol"
  ],
  "properties": {
    "after": {
      "type": "string",
      "description": "Optional opaque, exclusive event cursor for this symbol and tier."
    },
    "symbol": {
      "type": "string",
      "description": "Perpetual pair such as BTCUSDT."
    }
  },
  "additionalProperties": false
}
get_liquidation_census
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED. Hyperliquid exact liquidation census for one symbol. Aggregate price bands with forced long and short notional, stated coverage of venue open interest, snapshot identity and data age. A free API key receives a stored delayed snapshot. Pro keys and prepaid credits receive the current snapshot. No account rows. Separate from estimated liquidation walls. Descriptive census, not a forecast.

get_liquidation_census input schema
{
  "type": "object",
  "required": [
    "symbol"
  ],
  "properties": {
    "symbol": {
      "type": "string"
    }
  },
  "additionalProperties": false
}
desk_read
Advertised by the server during the last successful anonymous check.

FREE delayed keyless, live with paid access. Descriptive liquidation walls, cascade path, crowd and position distance, plus lines[] and text. GET /api/desk_read/symbols lists coverage. Unknown symbols return unknown_symbol and invalid numbers return invalid_number, with isError:true; live-only symbols on the free tier return symbol_requires_live_access with the free set in did_you_mean. The deprecated compatibility field verdict is wall_in_band, no_wall_in_band or thin_data, never null; proceed, reduce and skip are no longer emitted (contract change, SDK 0.4.1 docs). wall_in_band means the nearest wall on either side is under 1 percent from price, the same distance as flags.liq_distance_pct and the same test as flags.wall_within_1pct; thin_data means no price or no wall data. verdict_long and verdict_short apply that test to the losing-side wall (below for a long, above for a short) at the current price and assumed_leverage, or for your position. confidence is null with confidence_basis unscored until the public ledger scores verdicts. Compatibility metadata includes verdict_side, verdict_long, verdict_short, verdict_other, bias and confidence. The verdict, stop and target fields are deprecated compatibility fields, never advice. verdict is descriptive (wall_in_band, no_wall_in_band or thin_data) and reads the same wall set as the descriptive flags liq_distance_pct, wall_within_1pct and crowd_aligned, which replace it. stop_suggestion and target_suggestion mirror the deprecated plan and render on no customer surface.

desk_read input schema
{
  "type": "object",
  "required": [
    "symbol"
  ],
  "properties": {
    "ref": {
      "type": "string",
      "description": "Optional. Attribution tag."
    },
    "lang": {
      "enum": [
        "en",
        "zh",
        "ja",
        "ko",
        "es",
        "pt"
      ],
      "type": "string",
      "description": "Optional. Language of lines, text and act_label. Default en."
    },
    "side": {
      "enum": [
        "long",
        "short"
      ],
      "type": "string",
      "description": "Optional. Your position side."
    },
    "entry": {
      "type": "number",
      "description": "Optional. Your entry price."
    },
    "symbol": {
      "type": "string",
      "description": "e.g. BTCUSDT"
    },
    "leverage": {
      "type": "number",
      "description": "Optional. Your leverage."
    }
  }
}
hk_levels
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED (a free key works). Liquidation levels for one coin: the same hk.levels.v1 payload as REST GET /api/v1/levels/{symbol} — the nearest wall above and below the mark, every wall with price, side, usd, signed distance_pct, first_seen_ts, age_hours and an evidence label (validated, measured or unmeasured), the snapshot as_of and age_sec, and the public record link. A free key reads the 2 hour delayed snapshot on the free coins (one prepaid credit buys a live coin); a paid key reads the current snapshot. Descriptive, not a forecast.

hk_levels input schema
{
  "type": "object",
  "required": [
    "symbol"
  ],
  "properties": {
    "symbol": {
      "type": "string",
      "description": "Perp symbol, e.g. BTCUSDT."
    }
  },
  "additionalProperties": false
}
get_proof_summary
Advertised by the server during the last successful anonymous check.

FREE (keyless). The public proof summary: headline walk-forward precision with sample n + confidence lower bound, venues measured, per-venue feed-trust verdicts, and evidence pointers — the same aggregate hunterkiller.io/proof serves. Nothing per-level. Descriptive, not a forecast.

get_proof_summary input schema
{
  "type": "object",
  "properties": {}
}
get_magnet_levels_teaser
Advertised by the server during the last successful anonymous check.

FREE (keyless). BTC/ETH liquidation-cluster teaser: the top-3 strongest clusters per side (short-liq fuel above price, long-liq below) — the same published teaser data the anonymous free map and the public /learn pages serve, with the free-tier delay/scrub applied by construction. Descriptive positioning context, not a forecast. The full graded ladder (every level with reach-rate + n + CI, real-time, full symbol universe) requires an API key.

get_magnet_levels_teaser input schema
{
  "type": "object",
  "properties": {
    "symbol": {
      "type": "string",
      "description": "Optional: BTCUSDT or ETHUSDT (the free-teaser majors). Omit for both."
    }
  }
}
get_feed_trust_summary
Advertised by the server during the last successful anonymous check.

FREE (keyless). Summary of the public feed-trust reconciliation audit: per-venue verdicts on OUR OWN liquidation feed with match rate, coverage and assessable-run n — the same fields hunterkiller.io/proof/feed_trust publishes. A descriptive audit of data quality, not a trading signal.

get_feed_trust_summary input schema
{
  "type": "object",
  "properties": {}
}
get_liquidation_magnets
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED. Calibrated liquidation magnets for one symbol: per-horizon walk-forward empirical reach-rate (reach probability with sample n + Wilson lower bound) plus the magnet density histogram. Measured out-of-sample; descriptive, not a forecast. Unvalidated horizons are withheld, never served as numbers.

get_liquidation_magnets input schema
{
  "type": "object",
  "required": [
    "symbol"
  ],
  "properties": {
    "frame": {
      "type": "string",
      "description": "Optional horizon filter: 15m | 1h | 4h | 1D. Omit for all horizons."
    },
    "symbol": {
      "type": "string",
      "description": "Perp symbol, e.g. BTCUSDT."
    }
  }
}
get_reach_stats
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED. Walk-forward horizon reach probabilities for one symbol (out-of-sample backtest record): per-horizon up/down band reach with n + 95% CI. Descriptive history, not a forecast.

get_reach_stats input schema
{
  "type": "object",
  "required": [
    "symbol"
  ],
  "properties": {
    "symbol": {
      "type": "string",
      "description": "Perp symbol, e.g. ETHUSDT."
    }
  }
}
get_feed_trust
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED. Per-venue reconciliation verdicts for OUR OWN liquidation feed (rolling 7d): each venue's ingested events verified against an independent check, with assessable-run n shown. A descriptive audit of data quality, not a trading signal.

get_feed_trust input schema
{
  "type": "object",
  "properties": {
    "venue": {
      "type": "string",
      "description": "Optional venue filter (e.g. binance, bybit, okx, deribit). Omit for all."
    }
  }
}
get_accuracy_proof
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED. The published accuracy-proof aggregate (the same payload as the public /proof/stats): walk-forward precision with n, baselines, and window. Descriptive track record, not a forecast.

get_accuracy_proof input schema
{
  "type": "object",
  "properties": {}
}
get_engine_read
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED. The full /engine_read payload for a symbol+horizon as served to non-owner customers: the descriptive engine read plus horizons_meta with the per-horizon lane label (empirical = a walk-forward-validated measured cell is serving; modeled = the honest-null legacy lane) and the strict per-horizon gate_validated flag. The layer recipe is scrubbed by the standard non-owner egress. A descriptive structure read, not a forecast.

get_engine_read input schema
{
  "type": "object",
  "required": [
    "symbol"
  ],
  "properties": {
    "symbol": {
      "type": "string",
      "description": "Perp symbol, e.g. BTCUSDT."
    },
    "horizon": {
      "type": "string",
      "description": "Optional: 1h | 4h | 24h | 168h | 720h (default 1h)."
    }
  }
}
get_proof_anchors
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED. The public projection of the daily proof-anchor chain: each anchor day's date + content digest + chain hash (the append-only tamper-evidence ledger for the accuracy track record), with the keyless chain-integrity verification recipe. HMAC signatures and the attestation key are NOT included. Verification metadata — descriptive, not a forecast.

get_proof_anchors input schema
{
  "type": "object",
  "properties": {
    "limit": {
      "type": "integer",
      "description": "Optional: how many of the most recent daily anchors to return (default 30, max 200)."
    }
  }
}
get_decision_packet
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED. One composite evidence packet for a symbol: validated liquidation magnets, historical reach, contagion, derived venue structure and feed trust. Components fail independently and report honest unavailable states. Descriptive evidence, not a forecast. No order is placed.

get_decision_packet input schema
{
  "type": "object",
  "required": [
    "symbol"
  ],
  "properties": {
    "symbol": {
      "type": "string",
      "description": "Perp symbol, e.g. BTCUSDT."
    }
  }
}
prepare_trade
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED. Calculate position quantity, notional and optional gross reward-to-risk from user-supplied levels, with a supporting decision packet. Fees, funding and slippage are explicitly excluded. Descriptive preparation, not a forecast. Never signs or executes.

prepare_trade input schema
{
  "type": "object",
  "required": [
    "symbol",
    "direction",
    "entry",
    "invalidation",
    "risk_usd"
  ],
  "properties": {
    "entry": {
      "type": "number",
      "exclusiveMinimum": 0
    },
    "symbol": {
      "type": "string"
    },
    "target": {
      "type": "number",
      "exclusiveMinimum": 0
    },
    "risk_usd": {
      "type": "number",
      "exclusiveMinimum": 0
    },
    "direction": {
      "enum": [
        "long",
        "short"
      ],
      "type": "string"
    },
    "invalidation": {
      "type": "number",
      "exclusiveMinimum": 0
    }
  }
}
list_symbols
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED. The serveable perp symbol universe (the positive allowlist) — the same list as keyed REST GET /api/v1/symbols. Coverage metadata — descriptive, not a forecast.

list_symbols input schema
{
  "type": "object",
  "properties": {}
}
buy_credits
Advertised by the server during the last successful anonymous check.

API KEY REQUIRED (a free key works). Buy prepaid live Desk Read calls with a card: returns a Stripe Checkout URL to hand to your human. Packs: p10 ($10, 600 calls), p50 ($50, 3500), p200 ($200, 16000). With credits a free key gets live Desk Read, one credit per call. Balance: GET /api/agents/credits. This is a purchase action, not a trading signal.

buy_credits input schema
{
  "type": "object",
  "required": [
    "pack_id"
  ],
  "properties": {
    "pack_id": {
      "enum": [
        "p10",
        "p50",
        "p200"
      ],
      "type": "string"
    },
    "return_url": {
      "type": "string",
      "description": "Optional hunterkiller.io URL to return to."
    }
  }
}
feedback
Advertised by the server during the last successful anonymous check.

FREE, no key needed. Tell us what you needed and did not get; a human and our agents read every item. Missing field, wrong shape, unclear verdict, symbol we lack, latency, price: say it plainly. Max 2 KB, 5 items per hour per address. Never include keys or secrets; items that look like secrets are refused. Service feedback, not a trading signal.

feedback input schema
{
  "type": "object",
  "required": [
    "need"
  ],
  "properties": {
    "need": {
      "type": "string",
      "description": "What you needed and did not get (required)."
    },
    "tool": {
      "type": "string",
      "description": "Optional. The tool or endpoint this is about."
    },
    "symbol": {
      "type": "string",
      "description": "Optional. Perp pair this is about."
    },
    "contact": {
      "type": "string",
      "description": "Optional. Where a human can reply."
    },
    "context": {
      "type": "string",
      "description": "Optional. What you were trying to do."
    }
  }
}

Prompts

Not advertised

No prompt definitions available.

Resources

3 advertised

Data API catalog

Advertised by the server during the last successful anonymous check.

The machine-readable endpoint catalog (the same document /api/v1/catalog serves).

Data API catalog URI
hk://catalog

MIME type: application/json

Accuracy proof aggregate

Advertised by the server during the last successful anonymous check.

The published track-record aggregate (the same payload as public /proof/stats).

Accuracy proof aggregate URI
hk://proof

MIME type: application/json

Feed reconciliation verdicts

Advertised by the server during the last successful anonymous check.

Per-venue reconciliation audit of our own liquidation feed (public /proof/feed_trust).

Feed reconciliation verdicts URI
hk://feed-trust

MIME type: application/json

30-day availability

Aggregated from read-only remote checks; this is not a tool-execution or security guarantee.

Successful checks
1 of 1 (100%)
Average latency
2179 ms
95th percentile latency
2179 ms

Maintenance history

Public check and source-refresh events. Editorial and ownership records remain private.

  • · probe · passed · 2179 ms · metadata changed · 18 tool(s), 3 resource(s); warnings: Advertises resources/templates/list but does not implement it

Connect from your client

Each snippet registers the same endpoint; pick the client you use.

This server is operated by the named third party. Listing and endpoint checks are not an endorsement or a security audit. Review the operator's permissions, terms, and data handling before connecting. How listings are checked and corrected.

Claude Code

Run this once in a terminal. It registers the remote server. Use --scope project to write .mcp.json for the repo.
Claude Code configuration
claude mcp add --transport http hunter-killer-mcp https://hunterkiller.io/mcp

Codex

Run this once. Codex stores MCP servers in ~/.codex/config.toml (or .codex/config.toml in a trusted project).
Codex configuration
codex mcp add hunter-killer-mcp --url https://hunterkiller.io/mcp

Cursor

Merge this server into .cursor/mcp.json in your project, then enable hunter-killer-mcp in your MCP settings.
Cursor configuration
{
  "mcpServers": {
    "hunter-killer-mcp": {
      "url": "https://hunterkiller.io/mcp"
    }
  }
}

Cline

Merge this into cline_mcp_settings.json (Cline → MCP Servers → Configure).
Cline configuration
{
  "mcpServers": {
    "hunter-killer-mcp": {
      "type": "streamableHttp",
      "url": "https://hunterkiller.io/mcp",
      "disabled": false
    }
  }
}

VS Code

Merge this server into .vscode/mcp.json. Use MCP: List Servers to start hunter-killer-mcp and inspect its tools.
VS Code configuration
{
  "servers": {
    "hunter-killer-mcp": {
      "type": "http",
      "url": "https://hunterkiller.io/mcp"
    }
  }
}

Gemini CLI

Run this once. It writes .gemini/settings.json in the current project; pass --scope user for ~/.gemini/settings.json.
Gemini CLI configuration
gemini mcp add --transport http hunter-killer-mcp https://hunterkiller.io/mcp

Claude Desktop

Open Claude Settings > Connectors > Add custom connector, enter this server URL and enable it for the conversation. This is a manual remote-connector handoff, not a desktop JSON configuration.
Claude Desktop configuration
https://hunterkiller.io/mcp

Windsurf

Merge this into ~/.codeium/windsurf/mcp_config.json, then refresh MCP in Cascade. Windsurf remote servers use serverUrl, not url.
Windsurf configuration
{
  "mcpServers": {
    "hunter-killer-mcp": {
      "serverUrl": "https://hunterkiller.io/mcp"
    }
  }
}

Let your agent find its tools

Find servers like Hunter Killer MCP from your agent. Add Koskamo once to search the index for endpoints, access requirements, and client setup. Public discovery needs no key.