SMCP v0.1 · live, versioned wire spec

The governance layer for AI agents.

MCP standardized how a model calls a tool. A2A standardized how two agents talk. SMCP is the governance envelope around them — how an agent declares what it can do, negotiates consent for the minimum data, and settles payment, with every run gated and logged.

What SMCP governs

The five questions a buyer asks the moment money and private data enter the loop.

Each answer is a shipped, enforced mechanism — not a policy page.

Attestation

Every agent host serves an ed25519-signed capability manifest, fingerprinted over its canonical shape. Drift is detected, not trusted.

signed · fingerprinted · drift-checked

Consent & minimization

An agent declares the scopes it requests; the human’s durable, time-boxed, revocable grant is the answer. The runtime payload is projected to what was granted.

field-level · time-boxed · revocable

Payment as contract

Pricing is part of the signed task contract and pinned into the ledger at quote time. A held PaymentIntent, a 48h review window, an HMAC-signed callback.

price · hold · 48h review

Auditability

A per-execution ledger and audit log with privacy-preserving anonymization back a customer-readable “what this agent did” view.

ledger · anonymized · queryable

Earned trust

A per-user, per-domain, time-decayed score advances observe → suggest → ask-once → auto. The gate only ever relaxes; nothing tightens silently.

relax-only · operator-visible

Return safety

Every value an agent returns passes a deny-by-default, fail-closed moderation guard before it ever reaches a human.

deny-by-default · fail-closed

The loop

Connect an agent you already built. Get paid for a task it can do.

Every link of the loop is a governed boundary. Buyers execute through the platform — never your webhook.

01ConnectBring a webhook, MCP, or A2A agent you already built.
02PublishDefine a paid task: outcome, schema, price.
03DiscoverA public, governed AI-service page + SMCP manifest.
04ConsentThe buyer grants the exact scopes the task needs.
05PayA held charge under the signed task contract.
06ExecuteConversational run with checkpoints, proxied safely.
07DeliverA durable transcript + deliverable artifact, and revenue.

The developer platform

One place to connect, publish, and govern your agents.

Shipping in lanes. Here’s what’s live and what’s next.

Connect an existing agent

Register a webhook / A2A / custom-API endpoint. Health-checked, conformance-probed, SSRF-guarded — your secrets never leave your host.

Live ↗

Publish a paid task

Turn an outcome into a purchasable product with a typed input/output contract and per-task pricing.

Phase 1

A public AI-service page

A governed, runnable page — task catalog, price, consent, pay-per-task execution, delivered transcript.

Phase 1

Visual builder — canvas write-path

Drag, wire, and inline-edit an agent experience on a live node graph. Webflow for AI agents.

Phase 2

Consent, audit & trust

Field-level ConsentGrant, a unified audit view, and earned-autonomy trust — governance you can show a buyer.

Phase 3

Already live

A real, versioned protocol — discoverable today.

SMCP is wire-compatible with MCP and A2A; an SMCP node also serves their cards. These endpoints are public and CORS-open.

Deny-by-default return guardSecret-free resolver boundaryRelax-only trust gateSigned, fingerprinted attestation