Agent Collaboration & Consultation
How ARMES agents seamlessly consult other specialist agents mid-thread to deliver multi-disciplinary synthesis without context loss.
Cross-Disciplinary Intelligence
In a real company, an executive doesn't write code, and a software engineer rarely writes brand copy. Instead, team members consult with one another to combine their respective domain expertise.
In ARMES, your AI agents have the built-in capability to consult other user-owned agents mid-thread.
Plan Availability
Agent-to-agent consultation is available on Eco, Pro, and Ultra plans. Consultations only occur across agents you own and have installed in your workspace.
How Mid-Thread Consultation Works
When you are conversing with an agent (for example, ARMES GPT) and ask a question that bridges multiple domains:
User: "We are designing a new payment checkout page. Outline the user psychological friction points and provide a robust Stripe webhook handler in TypeScript."- Autonomous Identification: ARMES GPT recognizes that the prompt requires both consumer behavioral psychology and specialized backend engineering.
- Consultation Sub-Call: ARMES GPT seamlessly reaches out to the specialist:
- Target Teammate: Code Architect
- Question: "Provide a hardened, type-safe Stripe webhook handler in TypeScript with signature verification and idempotency handling."
- Internal Specialist Execution: The Code Architect processes the query using its own specialized system prompt and preferred model configuration.
- Synthesis & Citation: ARMES GPT integrates the Code Architect's code block into its overarching response, displaying a clear consultation badge:
[Consulted @code-architect for Stripe Webhook Handler]
Direct User Delegation (@Mentions)
You can also explicitly instruct an agent to consult a teammate using an plain language in the chat composer:
Draft the Q3 launch email. Please check with the wordsmith writer agent to review the copy for tone, and ask the code architect to check our data privacy claims.The active agent will coordinate with the requested specialists in sequence, presenting their findings in a structured, unified response.
Safety & Loop Prevention
To prevent runaway token consumption or infinite consultation loops:
- Sub-consultations are capped at 2 tiers deep (Agent A can consult Agent B, but Agent B cannot trigger an endless loop of sub-agents).
- All consultations are executed under your standard cost budgets and respect Zero Data Retention.