Your agents call tools.
Put identity in between.
Claude, Cursor, and custom agents reach MCP servers through the Monosign MCP Gateway — every tool call authenticated, authorized, budgeted, approved when destructive, and audited. Which MCPs exist, who connects to them, what can they call: answered.
- Tool-level authorization
- Human-in-the-loop approvals
- Full MCP audit trail
The agent stack gets a control plane.
The MCP Gateway extends Monosign’s agent-identity governance — kill-switch, on-behalf-of, budgets, approvals — to every tool call your agents make.
Allow-list the agent ecosystem
Agents reach only the MCP servers you registered, and only the tools you switched on. Discovery is filtered — clients never even see what they cannot call.
- Catalog of approved MCP servers, nothing else routable
- Per-tool enable / disable, with bulk operations
- Filtered tool discovery — hidden means invisible
- Destructive tools flagged and treated differently
Every tool call is authorized
Calls carry an identity — agent or human — verified via OAuth/OIDC and checked against permissions, conditions, and budgets before anything executes.
- OAuth / OIDC bearer authentication per call
- Per-tool permission requirements on top of the allow-list
- On-behalf-of: agent ∩ user permissions, per-user budgets
- Conditional access by IP, time, and day — plus rate limits
A human before the damage
Destructive tools pause for approval from the agent’s owner — in the inbox, by email, or by push — and a suspended agent stops at the gateway instantly.
- Human-in-the-loop approval for destructive calls
- Approve from Management inbox, email, or push
- Deny, expiry, and timeout all block the call
- Kill-switch enforcement at the gateway edge
Secrets stay out of the client
Upstream MCP credentials live in the vault and are injected just-in-time at the gateway. What flows through is inspected, and everything is on the record.
- Vaulted upstream credentials, JIT-injected
- Content inspection for secrets and sensitive data
- Full audit: identity, tool, verdict, per call
- Per-server call logs in the Management console
Three steps to governed agent tooling.
Register servers
Add the MCP servers your teams use to the catalog; the gateway becomes the only route to them.
Scope the tools
Enable tools per server, attach permission requirements, and mark the destructive ones for approval.
Govern the calls
Agents connect through the gateway with real identities — every call authorized, budgeted, and audited.
Ready to start managing
identities the right way?
Spin up a fully-loaded trial tenant in under five minutes. No credit card. No sales gate.