The Synapse Ecosystem

AI tools are smart but can’t see the user.

Synapse makes AI aware of human states. It integrates with tools like Claude Code and VS Code, managing interruptions and matching help.

Open Synapse

Consent gated
Measures the system
not the person
Signals stay on-device

Synapse Ecosystem

A layer, not another assistant

Synapse is not another copilot.

Your team keeps the agents and models it already uses. Synapse is the layer underneath delivered primarily as an MCP server supplying what those tools can’t see on their own.

Human-State Adaptation
Consent-gated HSI signals — focus, fragmentation, capacity — so AI tools can time and shape their help.
Organizational Memory
A “why” graph of decisions and context, built from the tools your org already works in.
Connectors
Org-scoped GitHub, Slack, and Notion connectors feed the work-signal side of the picture.

Works with the host’s agent loop and model over MCP nothing to replace.

The AI Effectiveness Index

Everyone measures AI. No one measures effectiveness.

Each case study follows one research question from hypothesis to finding, and where relevant, into what shipped. 2 active studies, more added as they progress.

AEI Dashboard

Works with the host’s agent loop and model over MCP — nothing to replace.

Two views, one boundary

Engineers see. Leaders see systems.

Engineers

Private self-insight: flow, focus fragmentation, context switching, and (with consent) how work demand lines up against capacity.

Leaders

Privacy-preserving team-health aggregates — how the system of work affects the team. Never individual scores, never rankings.

Synapse measures how work and human capacity interact. It measures the system, not the person and that boundary is architectural, not policy:

Read Privacy & Data

The consent gate executes on the device before serialization — an unconsented field is never created, not merely hidden.
Aggregates require a minimum cohort of 10 consented participants, enforced in the aggregation service, not the UI.
Opt-out is invisible to managers — absence is indistinguishable from cohort-floor suppression.
There is no individual-record endpoint for organizational roles, and no configuration that creates one.
No emotion inference in workplace deployments — by design and by policy.

Principles

The rules Synapse is built on.

Human State Aids Humans.
never surveillance, never scoring.
Consent Is Mandatory
opt-in, transparent, revocable at any time.
Local First
Physiological signals remain on-device; Synapse uses validated HSI output.
Adaptation Over Automation
adapt systems around engineers, don’t replace them.

Why Synapse

AI understands the task, not the moment.

Coding agents grasp code and builds, but not the humans. A 2,000-line refactor is reasonable but harmful under cognitive load. Interruptions cost time; refocusing takes about 23 minutes, impacting productivity.

FAQ

Frequently Asked Questions

Is this employee monitoring?No. Synapse measures the system of work, not the person. Leaders see team aggregates of at least 10 consented people, never individual scores or rankings.

Does Synapse read what people type?No. It uses consent-gated HSI signals such as focus, fragmentation, and capacity. Behavioral signals are timing and rhythm, never content.

Do we have to replace our AI tools?No. Your team keeps the agents and models it already uses. Synapse is the layer underneath, delivered mainly as an MCP server.

Does human-state data leave the device?Physiological signals stay on the device, and Synapse uses validated HSI output. The consent gate runs on the device, so an unconsented field is never created.

What does Synapse diagnose?Nothing. Synapse runs no diagnosis and no emotion inference in workplace deployments. It shows how work demand lines up against capacity.

No. Synapse measures the system of work, not the person. Leaders see team aggregates of at least 10 consented people, never individual scores or rankings.

No. It uses consent-gated HSI signals such as focus, fragmentation, and capacity. Behavioral signals are timing and rhythm, never content.

No. Your team keeps the agents and models it already uses. Synapse is the layer underneath, delivered mainly as an MCP server.

Physiological signals stay on the device, and Synapse uses validated HSI output. The consent gate runs on the device, so an unconsented field is never created.

Nothing. Synapse runs no diagnosis and no emotion inference in workplace deployments. It shows how work demand lines up against capacity.

Bring Synapse to your team.

Synapse is live for AI-first engineering teams. It plugs into the agents you already run over MCP, measuring the system, not the person.