What it is
A small shared task packet
SAE stays focused on task shape. It is deliberately narrower than a full agent framework or product stack.
Overview
SAE helps agent systems agree on the work before they try to do the work.
Structured Agent Exchange is an open specification for sharing the shape of agent work across A2A and non-A2A systems. It does not replace transport, discovery, or runtime layers. It gives systems a small, neutral packet for keeping task meaning clear across handoff.
Core fields
- Objective
- Constraints
- Deliverable
- Delegation hints
- Review posture
- Instruction frame
How it fits
Designed to ride over A2A
A2A handles connection, tasks, messages, artifacts, and extensions. SAE adds a stable, reviewable work packet above that layer.
Why it helps
Less drift across systems
SAE helps reduce ambiguity when work moves across different vendors, runtimes, or environments.
Start here
Reading path
Overview
The short explanation of what SAE is and why it exists.
README
Scope, framing, and spec index.
Quickstart
The fastest path to a first SAE packet.
Tutorial
Walk a fuzzy ask into a stable shared task.
SAE over A2A
How SAE layers over the A2A protocol.
Non-A2A translation
Adapters for HTTP, MCP-style hosts, and CLI workflows.
Wiki
Reference path, concepts, and supporting material.
Validation
How SAE is being tested
Benchmark bundle
Benchmark pack
SAE includes a reproducible benchmark bundle for comparing plain A2A, SAE over A2A, and a compiler-shaped SAE lane.
Open benchmark notesConformance
Schema and validation
Public schema, examples, and conformance notes define the base specification before wider adoption claims.
Open conformanceExternal scoring
Evaluation plan
The benchmark bundle includes outputs and an external evaluation plan for comparing SAE-shaped exchange against plain baseline exchange.
Open evaluation notes