Keryx

Research with Keryx

research

Research a question under a USDC creator-payment budget. Keryx selects sources, pays access tolls and weighted citation rewards on the configured Arc network, then returns qualified source excerpts and a receipt. Complete synthesis and per-assertion entailment remain unverified. This remote surface uses Keryx's treasury; anonymous research is sponsored, not caller-funded usage. Public research may send your question to our search provider. The source USDC budget is separate from model and search operating costs. For retained or repository bibliography use free paper_lookup with an exact identifier. An ordinary explicit exact-original bibliography request can also return a separate metadata-only record and references without a model call or creator payment; page status and full-paper evidence remain separate. New runs record remote MCP ingress. A verified ask-scoped API key attributes the run to its wallet; anonymous runs have no wallet owner. Editable client names establish no identity or spending authority.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "question": "<question>"
    },
    "name": "research"
  }
}

Input Schema

Name	Required	Description
`mode`	No	Research depth; default deep.
`model`	No	Optional Keryx model catalog id, for example deepseek-chat.
`budget`	No	Maximum creator-payment budget in USDC; 0 allows free sources without purchases or rewards. Clamped to the caller's tier.
`language`	No	Language of generated labels only; the original question and answer are unchanged. Omitted labels follow qualified recorded answer language when available; historical and unqualified answer language remains unknown.
`question`	Yes	Research question.
`scholarly`	No	Opt in to bounded Crossref/arXiv paper discovery; sends the question to those providers.
`reviewFirst`	No	Review-first requires the authenticated live browser; this adapter refuses true.
`researchMode`	No	Unsupported alias; omit this field and use mode for research depth.
`responseMode`	No	Unsupported response override; omit this field.

Schema Changelog

Changes observed during successful MCP inspections.

Changed2 schema fields changedNov 7, 2026
- addedInput schema / properties / researchMode / description
```
Added value: +"Unsupported alias; omit this field and use mode for research depth."
```
- addedInput schema / properties / responseMode / description
```
Added value: +"Unsupported response override; omit this field."
```

Changed1 schema field changedNov 1, 2026

changedInput schema / properties / language / enum

Previous value: -[
-  "en",
-  "vi",
-  "pt",
-  "pt-BR",
-  "es"
-]New value: +[
+  "en",
+  "vi",
+  "pt",
+  "pt-BR",
+  "es",
+  "fr",
+  "zh-Hans",
+  "de",
+  "ja",
+  "ko",
+  "hi",
+  "id",
+  "ru",
+  "ar"
+]

Changed1 schema field changedOct 13, 2026

changedInput schema / properties / language / description

Previous value: -"Language of generated labels only; the original question and answer are unchanged. Answer language is not recorded by this producer."New value: +"Language of generated labels only; the original question and answer are unchanged. Omitted labels follow qualified recorded answer language when available; historical and unqualified answer language remains unknown."

Changed1 schema field changedOct 12, 2026

addedInput schema / properties / language

Added value: +{
+  "description": "Language of generated labels only; the original question and answer are unchanged. Answer language is not recorded by this producer.",
+  "enum": [
+    "en",
+    "vi",
+    "pt",
+    "pt-BR",
+    "es"
+  ],
+  "type": "string"
+}

Changed3 schema fields changedOct 11, 2026

addedInput schema / properties / researchMode
```
Added value: +{
+  "not": {}
+}
```
addedInput schema / properties / responseMode
```
Added value: +{
+  "not": {}
+}
```

addedInput schema / properties / reviewFirst

Added value: +{
+  "const": false,
+  "description": "Review-first requires the authenticated live browser; this adapter refuses true.",
+  "type": "boolean"
+}

Changed3 schema fields changedOct 5, 2026

changedInput schema / properties / budget / description

Previous value: -"Maximum creator-payment budget in USDC; clamped to the caller's tier."New value: +"Maximum creator-payment budget in USDC; 0 allows free sources without purchases or rewards. Clamped to the caller's tier."

removedInput schema / properties / budget / exclusiveMinimum
```
Removed value: -0
```
addedInput schema / properties / budget / minimum
```
Added value: +0
```

Changed2 schema fields changedOct 1, 2026

addedInput schema / properties / mode

Added value: +{
+  "description": "Research depth; default deep.",
+  "enum": [
+    "quick",
+    "deep"
+  ],
+  "type": "string"
+}

addedInput schema / properties / scholarly

Added value: +{
+  "description": "Opt in to bounded Crossref/arXiv paper discovery; sends the question to those providers.",
+  "type": "boolean"
+}

First observedJul 27, 2026

TDQS

A4.4/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Beyond the annotations (readOnlyHint=false, openWorldHint=true, idempotentHint=false), the description discloses that the call spends real USDC from a budget separate from model and search operating costs, that the question may be sent to an external search provider, and that each new run is a fresh non-idempotent event recorded as remote MCP ingress. It also states who funds the run — Keryx's treasury for anonymous usage versus a verified ask-scoped API key that attributes the run to a wallet — which the annotations cannot convey.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness3/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The strong opening is front-loaded, but the text then sprawls into roughly ten sentences mixing cost, privacy, identity attribution, and disclaimers. Lines like "Editable client names establish no identity or spending authority" and the repeated "remain unverified/remain separate" phrasing dilute the actionable core and could be compressed into two or three sentences without losing meaning.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a nine-parameter, open-world, money-spending tool with no output schema, the description covers cost model, privacy exposure, attribution, result shape (excerpts plus receipt), and what is explicitly not verified (synthesis, per-assertion entailment). Remaining gaps are secondary — expected latency, failure or refund behavior when a purchase fails, and what a receipt contains.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema coverage is 100%, so the schema already documents all nine parameters and the baseline is 3. The description adds genuine meaning on top, notably that the source USDC budget is separate from model and search operating costs and that the budget is clamped to the caller's tier, both of which affect how an agent should set budget.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The first two sentences give a specific verb and resource ("Research a question") plus the mechanism: Keryx selects sources, pays access tolls and citation rewards on Arc, and returns qualified excerpts and a receipt. It also explicitly distinguishes itself from the sibling paper_lookup, so an agent can route without opening either schema.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

It names one concrete alternative with its condition ("For retained or repository bibliography use free paper_lookup with an exact identifier") and explains that a metadata-only bibliography path exists without a model call or creator payment. What is missing is general when-not-to-use guidance for the paid run itself, e.g., when quick vs. deep is warranted or when to prefer evidence_procurement or draft_check.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All Keryx tools Inspect

Try in Browser

One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.