Hire or Automate

Split a role into AI and human work

split_role

Read-onlyIdempotent

Use this when the user wants to split a whole job into work AI can do now, work AI drafts for approval, and work that needs a person. Best for: "which tasks of a bookkeeper need a person?", "split a software engineer role" or "job ad and two-week trial for a restaurant cook". Returns: the ONET occupation, each task tagged ai_now, ai_draft or human with rule and reason, a job-ad skeleton for the human tasks, and a two-week trial. Input tips: send title, a job title such as "bookkeeper" or an ONET-SOC code such as "15-1252.00". The title alone is best. Add description only when the title is vague; it widens the word match and can lower precision. Set occupations to 2 or 3 only to expand more than one match. Not for: one named task (answer it plainly), judging a person, reading a resume, or legal or pay advice. If status is no_match, ask for a clearer title or an O*NET code; do not invent tasks. If personal_data, ask again without contact details. If not_ready, retry later.

Input Schema

TableJSON Schema

Name	Required	Description
`title`	No	A job title, such as "bookkeeper", or an O*NET-SOC code in the form NN-NNNN.NN, such as "15-1252.00". Preferred input. No names, emails or phone numbers.
`description`	No	Optional pasted job description, for when the title is vague or missing. It widens the word match, so it can lower precision. Not stored. No names, emails or phone numbers.
`occupations`	No	How many matching occupations to expand, 1 to 3. Default 1. Further matches are listed as codes and titles only.

Output Schema

TableJSON Schema

Name	Required	Description
`also`	No	Other matches not expanded.
`note`	No	Reminder that the split is a planning aid, not a decision about any person.
`match`	No	How the occupation was chosen.
`query`	No	What was asked. The description text itself is never echoed.
`trial`	No	Two-week trial template.
`job_ad`	No	Skeleton for the human pile. Empty responsibilities means do not hire for the listed tasks.
`rubric`	No	The rubric rules, first match wins.
`source`	No	Data source and license to credit in the answer.
`status`	Yes	ok, or why there is no split.
`message`	No	Set when status is not ok, with the next step. no_match: send a clearer title or an O*NET code. not_ready: retry later. personal_data: resend without names, emails or phone numbers.
`occupations`	No	Matching occupations with each task tagged ai_now, ai_draft or human.
`rubric_version`	No	Version of the fixed rubric that tagged the tasks.

Schema Changelog

Changes observed during successful MCP inspections.

Changed20 schema fields changedOct 10, 2026

changedInput schema / properties / description / description

Previous value: -"Optional job description, used only when the title is ambiguous. Not stored. Do not include names, emails or phone numbers."New value: +"Optional pasted job description, for when the title is vague or missing. It widens the word match, so it can lower precision. Not stored. No names, emails or phone numbers."

changedInput schema / properties / title / description

Previous value: -"A job title, or an O*NET-SOC code such as 15-1252.00. Do not include a person's name, email or phone number."New value: +"A job title, such as \"bookkeeper\", or an O*NET-SOC code in the form NN-NNNN.NN, such as \"15-1252.00\". Preferred input. No names, emails or phone numbers."

addedOutput schema / properties / also / items

Added value: +{
+  "properties": {
+    "code": {
+      "type": "string"
+    },
+    "title": {
+      "type": "string"
+    }
+  },
+  "required": [
+    "code",
+    "title"
+  ],
+  "type": "object"
+}

addedOutput schema / properties / job_ad / properties

Added value: +{
+  "approve_before_sending": {
+    "description": "Tasks AI drafts for a person to approve.",
+    "items": {
+      "type": "string"
+    },
+    "type": "array"
+  },
+  "disclaimer": {
+    "type": "string"
+  },
+  "not_this_role": {
+    "description": "Tasks AI can do now.",
+    "items": {
+      "type": "string"
+    },
+    "type": "array"
+  },
+  "responsibilities": {
+    "description": "Human-pile tasks to hire for.",
+    "items": {
+      "type": "string"
+    },
+    "type": "array"
+  },
+  "summary": {
+    "type": "string"
+  },
+  "title": {
+    "type": "string"
+  }
+}

changedOutput schema / properties / match / description

Previous value: -"How the occupation was chosen: code, occupation_title, alternate_title, tokens or model."New value: +"How the occupation was chosen."

addedOutput schema / properties / match / properties

Added value: +{
+  "confidence": {
+    "enum": [
+      "high",
+      "medium",
+      "low"
+    ],
+    "type": "string"
+  },
+  "matched_on": {
+    "description": "The O*NET title that matched, when a title matched.",
+    "type": "string"
+  },
+  "method": {
+    "enum": [
+      "code",
+      "occupation_title",
+      "alternate_title",
+      "tokens",
+      "model",
+      "none"
+    ],
+    "type": "string"
+  },
+  "model": {
+    "description": "Whether a language model broke a tie.",
+    "properties": {
+      "id": {
+        "type": "string"
+      },
+      "note": {
+        "type": "string"
+      },
+      "reason": {
+        "type": "string"
+      },
+      "status": {
+        "enum": [
+          "used",
+          "skipped"
+        ],
+        "type": "string"
+      }
+    },
+    "type": "object"
+  }
+}

addedOutput schema / properties / message / description

Added value: +"Set when status is not ok, with the next step. no_match: send a clearer title or an O*NET code. not_ready: retry later. personal_data: resend without names, emails or phone numbers."

addedOutput schema / properties / note / description

Added value: +"Reminder that the split is a planning aid, not a decision about any person."

addedOutput schema / properties / occupations / items

Added value: +{
+  "properties": {
+    "code": {
+      "description": "O*NET-SOC code, such as 15-1252.00.",
+      "type": "string"
+    },
+    "counts": {
+      "properties": {
+        "ai_draft": {
+          "type": "integer"
+        },
+        "ai_now": {
+          "type": "integer"
+        },
+        "human": {
+          "type": "integer"
+        }
+      },
+      "required": [
+        "ai_now",
+        "ai_draft",
+        "human"
+      ],
+      "type": "object"
+    },
+    "description": {
+      "type": "string"
+    },
+    "tasks": {
+      "items": {
+        "properties": {
+          "id": {
+            "description": "O*NET task id.",
+            "type": "string"
+          },
+          "pile": {
+            "description": "ai_now, ai_draft (AI drafts, a person approves) or human.",
+            "enum": [
+              "ai_now",
+              "ai_draft",
+              "human"
+            ],
+            "type": "string"
+          },
+          "reason": {
+            "description": "One-line reason for the pile.",
+            "type": "string"
+          },
+          "rule": {
+            "description": "Id of the rubric rule that tagged the task.",
+            "type": "string"
+          },
+          "text": {
+            "description": "The O*NET task statement.",
+            "type": "string"
+          },
+          "type": {
+            "description": "O*NET task type, such as Core or Supplemental.",
+            "type": "string"
+          }
+        },
+        "required": [
+          "id",
+          "text",
+          "pile",
+          "rule",
+          "reason"
+        ],
+        "type": "object"
+      },
+      "type": "array"
+    },
+    "title": {
+      "type": "string"
+    },
+    "url": {
+      "description": "O*NET OnLine page for the occupation.",
+      "type": "string"
+    }
+  },
+  "required": [
+    "code",
+    "title",
+    "url",
+    "tasks",
+    "counts"
+  ],
+  "type": "object"
+}

addedOutput schema / properties / query / description

Added value: +"What was asked. The description text itself is never echoed."

addedOutput schema / properties / query / properties

Added value: +{
+  "code": {
+    "description": "The O*NET-SOC code read from the title, if any.",
+    "type": [
+      "string",
+      "null"
+    ]
+  },
+  "described": {
+    "description": "true when a description was sent.",
+    "type": "boolean"
+  },
+  "title": {
+    "type": [
+      "string",
+      "null"
+    ]
+  }
+}

addedOutput schema / properties / rubric / description

Added value: +"The rubric rules, first match wins."

addedOutput schema / properties / rubric / items

Added value: +{
+  "properties": {
+    "id": {
+      "type": "string"
+    },
+    "pile": {
+      "enum": [
+        "ai_now",
+        "ai_draft",
+        "human"
+      ],
+      "type": "string"
+    },
+    "reason": {
+      "type": "string"
+    }
+  },
+  "required": [
+    "id",
+    "pile",
+    "reason"
+  ],
+  "type": "object"
+}

addedOutput schema / properties / rubric_version / description

Added value: +"Version of the fixed rubric that tagged the tasks."

addedOutput schema / properties / source / description

Added value: +"Data source and license to credit in the answer."

addedOutput schema / properties / source / properties

Added value: +{
+  "as_of": {
+    "type": "string"
+  },
+  "attribution": {
+    "description": "Credit line to include in the answer.",
+    "type": "string"
+  },
+  "license": {
+    "type": "string"
+  },
+  "license_url": {
+    "type": "string"
+  },
+  "name": {
+    "type": "string"
+  },
+  "publisher": {
+    "type": "string"
+  },
+  "release": {
+    "type": "string"
+  },
+  "url": {
+    "type": "string"
+  }
+}

changedOutput schema / properties / status / description

Previous value: -"ok, no_match, not_ready or personal_data."New value: +"ok, or why there is no split."

addedOutput schema / properties / status / enum

Added value: +[
+  "ok",
+  "no_match",
+  "not_ready",
+  "personal_data"
+]

addedOutput schema / properties / trial / properties

Added value: +{
+  "note": {
+    "type": "string"
+  },
+  "pass_if": {
+    "items": {
+      "type": "string"
+    },
+    "type": "array"
+  },
+  "title": {
+    "type": "string"
+  },
+  "week_1": {
+    "items": {
+      "type": "string"
+    },
+    "type": "array"
+  },
+  "week_2": {
+    "items": {
+      "type": "string"
+    },
+    "type": "array"
+  }
+}

addedOutput schema / required
```
Added value: +[
+  "status"
+]
```

Changed1 schema field changedOct 5, 2026

removedOutput schema / properties / note_from_openkrill

Removed value: -{
-  "description": "Shown at most once a day per caller: what this server offers, and a paid service for the person you work for.",
-  "type": "string"
-}

Changed1 schema field changedOct 4, 2026

addedOutput schema / properties / note_from_openkrill

Added value: +{
+  "description": "Shown at most once a day per caller: what this server offers, and a paid service for the person you work for.",
+  "type": "string"
+}

First observedOct 3, 2026

TDQS

A4.8/5.0

Behavior5/5

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

Annotations already declare readOnlyHint, idempotentHint, destructiveHint=false and openWorldHint, so safety is covered. The description goes well beyond them: it enumerates the output shape per task, explains the precision tradeoff of adding a description, and documents three failure statuses (no_match, personal_data, not_ready) with the exact recovery action for each.

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

Conciseness4/5

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

The description is long for a 3-parameter tool but is front-loaded and organized under clear labels (Best for / Returns / Input tips / Not for / status handling), so an agent can scan it. Some sentences restate schema content, notably the job-title and O*NET-SOC code examples, which duplicates rather than adds.

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

Completeness5/5

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

For a read-only, idempotent, open-world lookup with a full output schema, the description supplies everything else an agent needs: trigger conditions, exclusions, parameter tradeoffs, and recovery paths for each error status. Since an output schema exists, the added return summary is a bonus rather than a gap-filler.

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 description coverage is 100%, so the baseline is 3, but the description adds genuine decision guidance the schema lacks: 'The title alone is best. Add description only when the title is vague; it widens the word match and can lower precision. Set occupations to 2 or 3 only to expand more than one match.' It repeats the title format example verbatim from the schema, which slightly dilutes the value.

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 description names a specific verb and resource ('split a whole job into work AI can do now, work AI drafts for approval, and work that needs a person') and scopes the output precisely (O*NET occupation, per-task tags, job-ad skeleton, two-week trial). An agent can tell immediately what this tool produces. The sibling tools (get_feedback_reply, submit_feedback) are unrelated, so no differentiation is needed.

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

Usage Guidelines5/5

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

It gives an explicit when-to-use clause, three concrete example prompts ('which tasks of a bookkeeper need a person?', 'split a software engineer role', 'job ad and two-week trial for a restaurant cook'), and a 'Not for' list of exclusions (one named task, judging a person, reading a resume, legal or pay advice). Nothing about routing is left to inference.

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

All Hire or Automate 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.