Marsell

Experiment start

experiment_start

Start a test with a hypothesis: two or more variants of one variable on one surface. Paid tests are only allowed when the playbook allows spend. Variants become work; the numbers come back through outcomes. Returns the test's id; experiments_list shows it running and, once decided, its winner. It needs an act key.

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": {
      "name": "<name>",
      "hypothesis": "<hypothesis>",
      "variable": "<variable>",
      "surface": "<surface>",
      "variants": "<variants>",
      "metric": "<metric>",
      "minSample": "<minSample>"
    },
    "name": "experiment_start"
  }
}

Input Schema

Name	Required	Description
`name`	Yes	A short name for the test.
`metric`	Yes	rate: the share who convert (clicks or sign-ups out of those who saw it); mean: an average amount per person (money, time).
`project`	No	The project's id or exact name. Optional when the account has one project.
`surface`	Yes	Where it runs: paid (ads, only when the playbook allows spend), organic (the business's own posts), email or page (a landing page).
`variable`	Yes	The one thing that differs between the variants: a headline, a picture, a price, a send time.
`variants`	Yes	Two or more variants; the one keyed control, else the first, is the control.
`minSample`	Yes	People each variant needs before the test may be decided; at least 30.
`hypothesis`	Yes	What you expect and why, in one sentence.

Schema Changelog

Changes observed during successful MCP inspections.

First observedNov 4, 2026

TDQS

A4.1/5.0

Behavior4/5

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

Annotations only say this is a non-destructive, closed-world write; the description adds substantive context beyond that — the spend-permission gate, that variants materialize as work items, that results arrive via outcomes, and that an act key is required. It does not cover failure modes or whether repeated calls create duplicate tests, which keeps it from a 5.

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?

Four short sentences, front-loaded with the definition, then constraints, then the return value and the auth requirement. Dense and mostly waste-free, though 'Variants become work' is terse domain jargon that an unfamiliar agent may not parse.

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?

With 7 required parameters and no output schema, the description does the necessary work: it states the return value (the test's id), the credential requirement (act key), and the spend precondition. Missing only error/duplicate-handling behavior for a mutation tool with substantial required inputs.

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

Parameters3/5

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

Schema coverage is 100%, so every parameter already carries its own explanation and the baseline is 3. The description reinforces the variable/surface constraint and the paid-surface restriction, but adds no format, default, or edge-case guidance (for example control-variant selection) beyond what the schema states.

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?

States a specific verb and resource ('Start a test') and immediately narrows the scope: two or more variants of one variable on one surface. That phrasing also distinguishes it from the read-only sibling experiments_list, which is named later as the way to observe the test.

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?

Gives a real conditional: paid tests are only allowed when the playbook allows spend, which tells the agent when this call will be rejected. It also points to experiments_list for monitoring and winner reporting, so the alternative is named. It stops short of stating explicit when-not conditions beyond the paid-spend gate.

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

All Marsell 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.