Skip to main content
Glama

start_duel

Run an artifact-building duel: two persona teams answer one brief, judges score both against a rubric, and the losing side files dissents; returns winner, tally, verdicts and archive id.

Instructions

Run a duel and return its result: two teams each build an artifact for the same brief (a web page, code, a Mermaid diagram or a text document), judges assess both against a rubric, and the losing side files dissents. The call runs the whole duel, which can take several minutes and sends a progress notification on every turn, artifact and verdict; the host's request timeout must be long enough or reset on progress. No room is opened. The result has the duel id, winner, tally, each team's artifact title, kind and ok flag, the judges' verdicts, the dissents, a usage summary and the archive id (duels/.json, replayable with roundtable duel --replay). It never includes artifact source.

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": {
      "topic": "<topic>",
      "team_a": "<team_a>",
      "team_b": "<team_b>",
      "judges": "<judges>"
    },
    "name": "start_duel"
  }
}

Input Schema

NameRequiredDescriptionDefault
kindNoWhat the teams build: web (a self-contained index.html, inline CSS, optional inline JS; default), code (source files, shown as read-only text and never executed), diagram (Mermaid source) or text (markdown or plain text).
modelNo
seatsNoOptional assignments keyed by exact persona identifier: {"designer":{"provider":"anthropic","model":"sonnet"}}. Each entry accepts agent, provider (a provider or explicit runtime name) and model.
topicYesThe brief both teams answer, for example a landing page for a product, a module with a given interface, or a written policy.
judgesYesJudge personas, 1 to 5.
roundsNo1 (default) judges once; 2 lets a team the judges did not approve revise once and keeps the original tally.
team_aYesTeam A personas, 1 to 4. The first writes the concept, the last builds, anyone between critiques the concept. A one-person team does both.
team_bYesTeam B personas, 1 to 4. Same roles as team_a. No persona may appear twice across both teams and the judges.
criteriaNoJudging rubric. Default: fit to the brief, quality, usability and accessibility, and risk or safety within the judge's own remit.
providerNoOmit or use auto for provider judgment (the router picks a healthy provider per seat and reports why in a routing block); otherwise the chosen provider is used as given. Options: auto; anthropic; openai; together (Together AI — any Together model id, e.g. zai-org/GLM-5.3 or Qwen/...); replicate (owner/name, owner/name:version, or https://replicate.com/owner/name); ollama (local, only when reachable); openai_compatible (any /v1/chat/completions server — set OPENAI_COMPATIBLE_BASE_URL or pass base_url).
connectorsNoOptional names of configured MCP connectors (roundtable config set connector) that every seat may call as tools during its turn. Only read-only tools are offered unless the connector's config lists more; tool results are untrusted data and are scrubbed. Seats on providers without tool support (Replicate) speak without them. An unknown connector is an error before any provider call.
page_checkNoCheck web pages in local headless Chrome when available (default true). Set false to skip.
max_secondsNoTime limit for the whole duel before it moves to closings. Default 600, plus 120 per member beyond two on the larger team. Hosts need a request timeout longer than this, or one that resets on progress notifications.
turn_timeout_sNoSeconds allowed for each turn. Default 90 (builds get at least 240).
seat_connectorsNoOptional per-seat limits, used with connectors: a map from a seat (agent) name to the subset of connectors that seat may call, e.g. {"ada": ["figma"], "bob": ["jira"]}. A seat not listed gets all of the session's connectors; an empty list gives that seat none. Every connector named must also be in connectors, and every seat must be at the table; otherwise the call fails before anything starts.

Schema Changelog

Changes observed during successful MCP inspections.

  1. Addedv2.0.0

TDQS

A4.4/5.0
Behavior5/5

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

Annotations already declare non-read-only, non-idempotent, open-world, non-destructive, but the description adds the operationally critical facts an agent cannot infer: multi-minute runtime, a progress notification on every turn/artifact/verdict, and the host request-timeout requirement (must exceed max_seconds or reset on progress). It also states what the result contains and, importantly, what it never contains (artifact source).

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?

Front-loaded with purpose, then runtime caveat, then the room exclusion, then the return shape and its exclusion — a sensible information order with no filler sentences. The first sentence is dense with nested parentheticals, which slightly slows scanning, but every clause earns its place given the absence of an output schema.

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 heavy 15-parameter, nested-object, non-idempotent tool with no output schema, the description supplies the missing pieces: duration, timeout/pagination-style progress behavior, and a full accounting of the return payload including the archive id and replay path. An agent can decide to call it and configure its host correctly without opening the schema.

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 description coverage is 93%, so nearly every parameter is already documented in the input schema, including kind, seats, rounds, team composition, criteria, connectors and timeouts. The description's mention of artifact kinds and rubric judging restates rather than extends the schema, and adds no syntax or constraint detail of its own. Baseline 3 applies.

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 opening sentence names a specific verb and resource ('Run a duel and return its result') and then describes the full mechanism — two teams build against one brief, judges score against a rubric, the loser files dissents. 'No room is opened' distinguishes it from the room-opening siblings like start_meeting without needing to inspect their schemas.

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?

Usage context is explicit: this runs the entire duel end-to-end in one call, taking several minutes, and it opens no room — the negative statement implicitly steers an agent toward the meeting siblings when a room/persistent session is wanted. It stops short of naming an explicit alternative or stating when-not to use it beyond that, so it is clear but not fully routing.

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