Skip to main content
Glama

Run a swarm

run_swarm

Have several AI agents on the user's own computers work on a question at once, through Toskar, and answer with what they found together.

Instructions

Have several AI agents on the user's own computers work on a question at once, through Toskar, and answer with what they found together. split divides the job into parts the agents do at once; race has agents on different models each answer the whole question, and the best answer wins. It takes longer than ask_local_ai, saves the agents' chats in Toskar, and returns the answer and a link to the coordinator's chat. Agents may read the web or the user's connected knowledge, but they don't change anything.

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": {
      "prompt": "<prompt>"
    },
    "name": "run_swarm"
  }
}

Input Schema

NameRequiredDescriptionDefault
modeNoOptional. split (the default) or race; leave out for the profile's.
promptYesThe question or job, with any context it needs.
profileNoOptional. The id of the Toskar profile to run it with; leave out for the user's default.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Addedv1.14.0

TDQS

A4.3/5.0
Behavior4/5

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

Beyond the annotations it discloses latency ('takes longer than ask_local_ai'), the side effect that explains readOnlyHint=false ('saves the agents' chats in Toskar'), the return payload ('the answer and a link to the coordinator's chat'), and the blast radius ('they don't change anything'). The only soft spot is that 'agents may read the web' sits in tension with openWorldHint=false, though that hint plausibly scopes the tool's own operation rather than the sub-agents' downstream access, so it reads as a hint tension rather than a clear contradiction.

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?

Three dense sentences, front-loaded with the purpose before the mode taxonomy and the cost/side-effect notes. Every clause carries information; the mode explanation is the longest part but is doing real work rather than padding.

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 three-parameter tool with no output schema, the description covers what it produces (combined answer plus coordinator chat link), how long it takes, and what it will not do. Remaining gaps — failure behavior, whether agents can be partially unresponsive, any limit on fan-out — are second-order for a definition of this size.

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 enum values and parameter roles are already documented; the description nevertheless adds genuine meaning by explaining what split and race actually do and when each wins, which the bare enum description does not convey. It adds no further guidance on 'profile' beyond the schema, but that is a minor omission.

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 sentence names a concrete verb and resource ('have several AI agents ... work on a question at once, through Toskar') and states the output ('answer with what they found together'). It also explicitly contrasts with the sibling ask_local_ai ('It takes longer than ask_local_ai'), so an agent can distinguish the two 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 gives real selection guidance between the two modes ('split divides the job into parts ... race has agents on different models each answer the whole question, and the best answer wins') and frames the cost tradeoff versus ask_local_ai. What it lacks is an explicit 'use ask_local_ai instead when you only need one model / need a fast answer' exclusion, so the choice is implied rather than stated.

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