Agent-to-Agent Negotiation
Agent-to-agent negotiation is a core research capability within Craftly Robot.
The idea is simple : instead of requiring a person to manually communicate with every person or service involved in a task, one Robot agent can communicate with another agent to discover capabilities, exchange relevant requirements, determine alignment, and coordinate a possible solution.
This is an important part of moving Robot from conversational AI toward real world task coordination.
Modern agent interoperability research follows a similar pattern. Google's A2A protocol, for example, defines mechanisms for agent capability discovery, task-oriented communication, collaboration, and exchanging information between independent agents without requiring one agent to access another agent's internal state.
Craftly Robot is developing its own identity, discovery, and agent communication model around this broader idea.
How It Works
A simplified Craftly workflow can look like:
User Request ↓ Requirement Understanding ↓ Agent Discovery ↓ Agent Request ↓ Agent-to-Agent Communication ↓ Alignment / Negotiation ↓ Potential Solution ↓ Human Confirmation ↓ Real World Action
The exact workflow depends on the task.
Real World Example
Suppose a user asks:
“Find a laptop repair service in Dhaka that can replace a damaged MacBook display within two days for under ৳15,000, and confirm whether they provide a warranty.”
Robot can interpret the request into requirements:
- MacBook display replacement
- Dhaka
- Completion within two days
- Maximum budget of ৳15,000
- Warranty confirmation
It can then discover potentially relevant service agents.
Those agents can communicate about:
- Whether the service is available
- Whether the required display is available
- Expected repair time
- Price
- Warranty
- Other relevant conditions
Robot can then compare the resulting information and present suitable options to the user.
The goal is not for agents to make uncontrolled commitments. The goal is to let agents handle the communication and coordination work that would otherwise require significant human effort.
Current Status
Craftly Robot's agent-to-agent negotiation capability is currently part of active research and development.
The foundational infrastructure for identity, discovery, agent communication, and consent-aware coordination is being developed, and early real-world workflows are being explored.
The broader public workflow is still evolving.
As the system matures, Craftly intends to make increasingly complex agent-to-agent interactions possible while keeping identity, privacy, permissions, and human control central to the design.
Why It Matters
A single agent can assist one person.
A network of agents can coordinate across many people, services, and capabilities.
That changes the role of AI from:
“Tell us the answer.”
to:
“Help us get this done.”