ProtocolReferenceMCPA2A

gcf

Name: gcf
Rating: 3.0 (45 reviews)
Author: blackwell-systems

by blackwell-systems

AI-native compact wire format for lossless, token-efficient agent data exchange

Updated Sep 5, 2026

Stars

Forks

View on GitHub

Summary

Define a compact, AI-native wire format for structured data (GCF) that reduces token usage while keeping lossless round-trips across formats. Uses a schema-aware, compressed representation optimized for frontier LLMs to achieve 50–92% fewer tokens than JSON and high fidelity across 17 formats. Spec v3.2 documents stable encoding rules and compatibility targets for model inputs/outputs and agent-to-agent exchange. This aligns with the A2A Protocol Pattern.

Why It Matters

As agents exchange structured payloads, payload size and faithful serialization directly affect model costs and downstream trust signals. GCF makes agent communication more efficient and deterministic, reducing ambiguous JSON inference and lowering token overhead that can distort evaluations or inflate costs. For multi-agent trust and A2A evaluation, a stable, lossless wire format improves reproducibility of interactions and gives clearer signals when judging agent behavior or failures. This is supported by the Model Context Protocol (MCP).

When to Use

Teams building agent-to-agent protocols or A2A evaluation pipelines that need deterministic, token-efficient structured payloads. This fits well with the Agent Protocol for cross-agent integration.

Use Cases

Reduce model input costs by encoding structured agent messages with 50–92% fewer tokens than JSON
Ensure lossless payload round-trips for agent evaluation and reproducible A2A benchmarks
Standardize agent-to-agent payloads in MCP or custom A2A protocols to improve interoperability and debugging

Need implementation help?

Expert guidance available