Skip to main content
Glama

create_agent

Create a draft AI agent with configurable type, model, and visibility; get its UUID to add a persona and publish it.

Instructions

Create a draft agent and return its UUID. Add a persona with create_schema (persona_config), then publish_agent. Returns PLAN_LIMIT_REACHED when the tenant agent limit is reached. Read zihin://skills/criar-agente.

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>"
    },
    "name": "create_agent"
  }
}

Input Schema

NameRequiredDescriptionDefault
bioNoAgent description or biography
nameYesInternal agent name
tagsNoCategorization tags
typeNoAgent type; orchestrator can invoke other agents through invoke_agentassistant
metadataNoAdditional JSON metadata, such as {"department":"sales","priority":"high"}
llm_configNoModel configuration. model uses provider.model or auto; temperature ranges from 0 to 2; fallback_chain is an ordered model list.
visibilityNoAgent visibilityprivate
commercial_nameNoDisplay name shown to end users

Schema Changelog

Changes observed during successful MCP inspections.

  1. Changed14 schema fields changedv2.2.4
    • changedInput schema / properties / bio / description
      Previous value: -"Descrição/biografia do agente"New value: +"Agent description or biography"
    • changedInput schema / properties / commercial_name / description
      Previous value: -"Nome comercial (exibido ao usuário final)"New value: +"Display name shown to end users"
    • changedInput schema / properties / llm_config / description
      Previous value: -"Configuração do LLM. model: \"provider.modelo\" (ex: \"openai.gpt-4.1-nano\", \"anthropic.claude-sonnet-4-6\") ou \"auto\". temperature: 0-2. fallback_chain: cadeia de modelos ordenada."New value: +"Model configuration. model uses provider.model or auto; temperature ranges from 0 to 2; fallback_chain is an ordered model list."
    • changedInput schema / properties / llm_config / properties / context_policy / description
      Previous value: -"\"physical\" (default, também quando ausente) = sem corte de resultado de tool nem teto de 40k; o limite é a janela real do menor modelo da cadeia (turno que não cabe para com outcome CONTEXT_LIMIT). \"legacy\" = política antiga, só explícita e a ser removida."New value: +"physical (default) uses the smallest model context window in the chain without tool-result clipping or a 40k cap; an oversized turn stops with CONTEXT_LIMIT. legacy selects the deprecated policy and must be explicit."
    • changedInput schema / properties / llm_config / properties / fallback_chain / description
      Previous value: -"Cadeia ORDENADA de fallback, sem repetição. Max 5. Ex: [\"openai.gpt-4.1-mini\", \"anthropic.claude-haiku-4-5-20251001\"]"New value: +"Ordered fallback chain with no duplicates, up to five models. Example: [\"openai.gpt-4.1-mini\",\"anthropic.claude-haiku-4-5-20251001\"]"
    • changedInput schema / properties / llm_config / properties / max_iterations / description
      Previous value: -"Teto de iterações do loop por turno. O runtime clampa em 30."New value: +"Maximum loop iterations per turn; capped at 30 by the runtime"
    • changedInput schema / properties / llm_config / properties / max_tokens / description
      Previous value: -"Teto de tokens de SAÍDA por chamada (≠ janela de contexto, ≠ teto de gasto)"New value: +"Maximum output tokens per model call; distinct from the context window and spending cap"
    • changedInput schema / properties / llm_config / properties / model / description
      Previous value: -"\"auto\" (roteamento por tier) ou \"provider.modelo\" (ex: \"openai.gpt-4.1-nano\"). Consulte zihin://models."New value: +"Use auto for tier-based routing or provider.model, such as openai.gpt-4.1-nano. See zihin://models."
    • changedInput schema / properties / llm_config / properties / temperature / description
      Previous value: -"0=determinístico, 2=máximo criativo"New value: +"0 = deterministic; 2 = maximum creativity"
    • changedInput schema / properties / metadata / description
      Previous value: -"Metadados adicionais em JSON livre. Ex: { \"department\": \"vendas\", \"priority\": \"high\" }"New value: +"Additional JSON metadata, such as {\"department\":\"sales\",\"priority\":\"high\"}"
    • changedInput schema / properties / name / description
      Previous value: -"Nome interno do agente"New value: +"Internal agent name"
    • changedInput schema / properties / tags / description
      Previous value: -"Tags para categorização"New value: +"Categorization tags"
    • changedInput schema / properties / type / description
      Previous value: -"Tipo do agente (orchestrator = invoca outros agentes via invoke_agent)"New value: +"Agent type; orchestrator can invoke other agents through invoke_agent"
    • changedInput schema / properties / visibility / description
      Previous value: -"Visibilidade do agente"New value: +"Agent visibility"
  2. Changed1 schema field changedv2.2.3
    • addedInput schema / properties / llm_config / properties / context_policy
      Added value: +{
      +  "description": "\"physical\" (default, também quando ausente) = sem corte de resultado de tool nem teto de 40k; o limite é a janela real do menor modelo da cadeia (turno que não cabe para com outcome CONTEXT_LIMIT). \"legacy\" = política antiga, só explícita e a ser removida.",
      +  "enum": [
      +    "physical",
      +    "legacy"
      +  ],
      +  "type": "string"
      +}
  3. First observedv0.1.0

TDQS

A4.3/5.0
Behavior4/5

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

With annotations already declaring readOnlyHint=false, destructiveHint=false, idempotentHint=false and openWorldHint=false, the description adds real context beyond them: the agent starts as a draft (not live), the UUID is returned, and a specific error code is surfaced on tenant plan limits. It does not say whether the call is retryable after the plan limit error or what other error modes exist.

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

Conciseness5/5

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

Four compact sentences, front-loaded with the action and its result, followed by the workflow, the error case, and a doc pointer. No sentence is redundant and none obscures the primary instruction.

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 an 8-parameter create tool with a nested config object and no output schema, the description still supplies the return value (UUID), the lifecycle state (draft), the follow-up steps needed for a usable agent, and the plan-limit error. That covers everything an agent needs to invoke this correctly.

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 100% and nested llm_config fields are individually documented, so the schema carries the parameter burden. The description adds no per-parameter meaning and only references persona_config indirectly through create_schema, so the baseline of 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?

States a specific verb and resource plus the artifact's lifecycle state: 'Create a draft agent and return its UUID.' The word 'draft' distinguishes this from publish_agent and clone_agent, and the mention of create_schema/publish_agent as follow-ups makes the boundary with those siblings explicit.

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 concrete sequencing guidance: create here, add a persona via create_schema (persona_config), then publish_agent. It also names the failure condition (PLAN_LIMIT_REACHED). It stops short of explicit when-not guidance or alternatives such as clone_agent for duplicating an existing agent.

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

Deploy Server

Other Tools