Run a bounded enterprise pilot
Start with approved tool names. Review the first report, then continue, adjust scope, or stop after Week 1.
See the four-week plan →MCP evidence for enterprise review
Know what your agents can call—and what you still do not know.
AgentGate turns an approved MCP tool inventory into a reviewable evidence record. Every source, failed check, correction, and unknown stays visible so your team can make a defensible decision.
No credentialsNo production accessOpen-source coreSelf-hostable
Where to start
Decision first. Product depth second.
Choose the shortest path that matches what you need to establish today.
Inspect a real deliverable
Open the sample report before sharing anything. Check how sources, failures, unmeasured work, and findings are presented.
Open the sample report →Use the open-source core
Review an inventory locally, inspect the public evidence index, or enforce a policy in your own environment.
Review your inventory →Four-week design partner pilot
Bring the list. Leave with a decision.
The customer keeps control of scope and access. AgentGate turns the agreed material into an evidence record, then asks whether the result changed a real governance or purchasing decision.
Review the mutual action planWeek 1Initial report and a continue, adjust, or stop gate.
Week 2Correct false positives, missed context, and evidence gaps.
Week 3Compare retained evidence when a relevant change exists.
Week 4Record the operational decision and the honest purchase path.
The free pilot does not become a subscription or public case study without a separate written agreement.
Evidence rules
The report shows where knowledge stops.
AgentGate is designed around evidence boundaries rather than optimistic defaults.
Required checks must finish
A result can be clean only when every required check completed. A skipped or crashed check remains incomplete.
Unmeasured stays unmeasured
Unavailable manifests, registry metadata, packages, or private configuration are visible gaps—not implied approvals.
History cannot be reconstructed
Retained snapshots make evidence changes visible. If no comparable snapshot exists, the report says so.
Open-source product
Inspect it before you trust it.
The current core can collect evidence, build an inventory, apply policy, retain change history, and place tool calls behind a runtime gateway. It is AGPL-3.0 and can be self-hosted.
Try the current productDiscoverProduce a local tool-name inventory without exporting credentials or environment variables.
AuditRecord registry, package, repository, scanner, and evidence coverage.
EnforceApply an explicit policy in CI or at a runtime gateway.
WatchRetain snapshots and show evidence changes that occur without a version change.
Current boundary: SSO/SAML, RBAC, multi-tenancy, enterprise retention controls, signed audit exports, and SIEM integrations are planned requirements, not available product capabilities today.
Verify the claims
Trust pages are part of the product.
Use the evidence, security, and data-handling pages to inspect what exists before contacting us.
Public evidence
Evidence index
Browse the current records, coverage, sources, and incomplete work.
Browse evidence → Verifiable controlsSecurity
Review release controls, vulnerability reporting, and explicit claim boundaries.
Review security → Data boundaryData handling
See what the service receives, what it records, and what remains on your systems.
Review data handling →A bounded first step
Start with evidence, not a platform commitment.
Review the sample, inspect the boundaries, then decide whether a four-week pilot is worth your team's time.
See the enterprise pilotOr email contact@智量.com · contact@xn--5kvo87g.com