Verify Hosted MCP

Do not just score MCP servers. Host the trusted ones behind a managed endpoint with validation, secrets, egress controls, releases, rollback, trust badges, and audit events.

Hosted endpoint

/hosted/{namespace}/{name}/mcp

Remote MCP surface for activated hosted deployments.

Trust gate

TrustOps tier

Requires a verified publisher claim plus paid hosted runtime feature access.

Release proof

score + snapshot

Each release carries the validation run and sandbox policy that allowed activation.

Control plane

SOURCE

Deploy from GitHub

Create a hosted runtime from a claimed server repository, source ref, and sandbox policy.

TRUST

Validate before activation

Attach the current Verify score, trust snapshot, validation run, and badge state to each release.

SECRETS

Runtime vault

Store header, token, env, or file secrets with encrypted payloads and redacted metadata.

SANDBOX

Egress allowlist

Deny outbound network access except the server endpoint and explicit allowlisted hosts.

RELEASES

Version and rollback

Promote releases, keep previous versions addressable, and roll back by release id.

AUDIT

Usage events

Emit deployment, secret, release, rollback, egress, upstream, and request usage events.

Who buys this

PUBLISHERS

Trusted runtime for claimed servers

Publishers who want a verified endpoint, trust badge, uptime monitoring, and release proof without running MCP infrastructure themselves.

PLATFORM TEAMS

Approved tools behind one endpoint

Agent platform teams that want trusted MCP servers hosted with consistent secrets, egress, release, and observability controls.

AI ADMINS

Runtime inventory with evidence

Enterprise admins who need hosted tools tied to policy decisions, audit events, and rollback-ready operational history.

How it connects to Gateway and Ledger

Hosted MCP turns a trusted server into an execution surface. Gateway decides whether a specific hosted call is allowed, and Ledger preserves the release, trust snapshot, mandate, outcome, and cost evidence.

GATEWAY

Decide before hosted execution

Gateway can require the hosted release id, trust snapshot, sandbox policy, and mandate before allowing an agent call.

LEDGER

Record runtime evidence

Ledger stores hosted deployment, release, rollback, egress, upstream, and request usage events next to policy and risk snapshots.

MANDATES

Bind authority to hosted calls

Mandates constrain which agents and users may call hosted tools, with budget, scope, and approval requirements evaluated at /v1/decide.

API surface

Method	Path	Purpose
`GET`	`/v1/hosting/capabilities`	Discover hosted runtime controls.
`GET`	`/v1/servers/{namespace}/{name}/hosting`	Inspect readiness, deployments, releases, secrets, and events.
`POST`	`/v1/servers/{namespace}/{name}/hosting/deployments`	Create or update a GitHub-backed hosted runtime deployment.
`POST`	`/v1/servers/{namespace}/{name}/hosting/deployments/{id}/secrets`	Store encrypted runtime secrets.
`POST`	`/v1/servers/{namespace}/{name}/hosting/deployments/{id}/releases`	Create and optionally activate a release.
`POST`	`/v1/servers/{namespace}/{name}/hosting/deployments/{id}/rollback`	Promote a previous release.
`POST`	`/hosted/{namespace}/{name}/mcp`	Serve the sandboxed hosted MCP endpoint.

Activation rule

Hosted runtime activation requires a healthy or degraded server, a remote endpoint, fresh validation, and a score at or above the hosting threshold. Deployments can be created while blocked, but releases stay pending until the trust gate passes.