Lowboy
Food cost intelligence for restaurants
Upload a distributor invoice. Get a variance report showing exactly where you're overpaying — line by line, dollar by dollar.
M87 Studio
Governance infrastructure
for AI systems.
Applied to the work that matters.
M87 builds systems where AI agents cannot act without cited, measurable proposals. The architecture runs audio analysis, kitchen operations, and coding agents — every claim grounded in signal or policy.
§ 02 · Thesis
What M87 builds.
AI agents propose and execute in the same step. That is the problem. Every public failure of agentic AI — a rogue commit, a deleted database, a hallucinated recommendation, a psychological inference presented as fact — traces back to the same structural flaw. Proposal and execution collapsed into a single decision.
M87's architecture separates them. Proposals are cited, measurable, and bound by governance. Execution runs through an authority-separated path. The model can propose anything. The system refuses to execute anything that cannot be grounded in signal or policy.
That architecture is the substrate. The products are what happen when you apply it to real work.
Live now.
Resonance.
Measurement-grounded proposals for audio.
Resonance translates what your audio is doing into cited proposals you approve or ignore. Measurements feed observations. Observations feed proposals. Every claim cites the signal behavior that motivated it. If a claim cannot be cited, it does not surface.
- 01Start a session. See what your audio is doing.
- 02Every proposal grounded in measurement. Every change receipted.
- 03Free tier available. Student pricing at $10 per month. Professional at $14 per month, unlimited.
M87 plugins coming. Built on the same signal-first system. Governed proposals surface inside the plugin, with approval at the parameter.
§ 04 · Product grid
The rest of the system.
Every M87 product instantiates the same governance architecture. One system applied to distinct domains — each product exists because the substrate works in that domain. Different work. Same enforcement.
Spine Lite
Governance for coding agents
Drop-in governance enforcement for Claude Code. Intercepts agent tool use at the runtime level. Blocks destructive operations through execution incapacity, not policy advisory.
Spine Pro
Distributed governance control plane
The full governance control plane above any AI coding agent. Ed25519-signed execution tokens, append-only receipt ledger, Gate authority separated from executor, four-key auth model. Fail-closed at every boundary. Docker-deployable.
Governed Swarm
Multi-agent governance substrate
Distributed governance for multi-agent workflows. Proposal envelopes, tamper-evident decision chains, adversarially validated across 328 test vectors and red-team phases. Public under BSL 1.1.
Resonance Bridge
Governed AI-assisted mixing
Session-level governance inside your DAW. The architecture reads your session, proposes cited adjustments grounded in signal behavior, and executes only with your approval. Works with DAW-native plugins. Works more deeply with M87 plugins, where every proposal surfaces inside the plugin itself with its rationale visible at the parameter.
§ 05 · Architecture
Why the architecture matters.
Most AI products treat governance as a feature you can add later. M87 treats it as the substrate everything else runs on.
Seven invariants govern every M87 product. Proposal is not execution. Authority is separated. The system fails closed on ambiguity. Nothing is complete without a verifiable artifact. Memory is structured, not accumulated. No model lock-in. Narrative never modifies runtime.
The Governance Kernel is available. Frozen schema, reference implementations in Python and JavaScript, a governance sandbox that falsifies the architecture under adversarial conditions. Licensed under BSL 1.1.
§ 06 · Who works with M87
Three ways in.
01
Producers and engineers
If you work with audio and you want a system that measures what your mix is doing and proposes changes you approve, start a session with Resonance.
Start a session→02
Educators and institutions
Twelve students, twelve instructors grading twelve different ways. Resonance gives every student the same measurement-grounded baseline, instructors a shared reference for critique, and programs a record of progression that doesn't depend on which ear graded it.
Institutional inquiry→03
Architects and engineering leaders
If you build with AI agents and you have reached the limits of policy-based guardrails, the governance architecture is available for consulting engagements and enterprise deployment.
Work with M87→§ 07 · Close
M87 Studio.
Founded in Portland, Oregon. Solo-built. Every M87 product instantiates the same seven invariants. Every claim in every product cites back to a measurable signal or a policy decision.
The thesis: AI agents become trustworthy when they structurally cannot act without proof. M87 builds the structures that make that true.
Governance is not a feature. It is the substrate.