Core Concepts
The Signal Gating Protocol sits between signal producers and the agents that consume their output. It is an executive-function layer: typed signals flow through composable gates, async channels transport them, and a mesh topology coordinates multi-agent workflows.
The neuroscience analogy
The brain does not hand every sensation to the cortex. The thalamus gates sensory input before it reaches higher regions. The prefrontal cortex decides what is currently relevant. Sparse coding keeps most neurons quiet, most of the time. Agents built on LLMs have no structural equivalent — SGP supplies one.
The four primitives
Signal
A typed, immutable event with lineage. A signal carries its payload, its schema, and a record of every gate it has passed through, so downstream agents can reason about provenance, not just content.
Gate
A composable predicate. Gates are combined with algebraic operators:
| Operator | Behavior |
|---|---|
chain(a, b, ...) | Apply gates in sequence; a signal must pass every gate. |
or(a, b) | A signal passes if either gate admits it. |
and(a, b) | A signal passes only if both gates admit it. |
invert(a) | A signal passes if the underlying gate rejects it. |
Built-in gates:
| Gate | Purpose |
|---|---|
filter(predicate) | Admit signals that satisfy a predicate. |
dedupe(key, ttl) | Drop duplicates within a TTL window. |
ttl(seconds) | Drop signals older than a deadline. |
throttle(rate, per) | Cap the rate at which signals pass. |
batch(size, timeout) | Group signals into batches. |
circuit_break(threshold, cooldown) | Trip open after a failure threshold; close after a cooldown. |
Channel
An async transport with backpressure. A channel is the only way signals cross process or network boundaries. A priority variant admits high-urgency signals ahead of the queue.
Mesh
A network topology over channels. A mesh expresses coordination between agents without hard-coding call graphs. Standard patterns:
| Pattern | Behavior |
|---|---|
| scatter / gather | Fan out to workers, collect responses. |
| map / reduce | Transform then aggregate. |
| race | Take the first responder. |
| branching | Route by predicate. |
Why this shape
Multi-agent systems built on LLMs fail in characteristic ways:
- Context collapse — every signal is forwarded to every agent until the window saturates with noise.
- Context rot — stale or redundant information displaces what matters.
- Cascade failures — one agent's hallucination propagates unchecked through downstream calls.
SGP's primitives exist to interrupt each of these failure modes at the protocol level, rather than through longer prompts and ad-hoc orchestration glue.
Next steps
- Try the Quickstart to compose a pipeline.
- Read the Specification for formal definitions.