Saturn is a secure MCP gateway between AI agents and the tools they call. It centralizes access, enforces fine-grained permissions, requires a person to approve sensitive actions, and keeps a complete audit trail in one console.

What is an MCP gateway?

MCP (Model Context Protocol) is the standard AI agents use to call tools. An MCP gateway sits in front of your MCP servers. Agents talk only to Saturn, which exposes the tools you choose, evaluates every call against policy, and forwards only the calls that are allowed.

Which AI clients work with Saturn?

Any MCP client, including Claude, ChatGPT, Cursor, Gemini, LM Studio, and Saturn App. Every agent uses one Saturn MCP endpoint. Existing MCP servers do not need code changes.

How is Saturn different from an API gateway?

An API gateway governs services by route and HTTP method. Saturn governs agents. It understands tools and arguments, can hold a call for human approval, and records which policy and which reviewer allowed it.

Do agents see API keys or OAuth tokens?

No. Upstream credentials stay in Saturn’s vault, sealed with ML-KEM-1024 post-quantum key encapsulation and AES-256-GCM. Agents receive only the result of an authorized call. APIs report whether a secret exists, never the secret itself.

Yes. Saturn App is a free, open-source desktop agent for macOS and Windows, MIT licensed. It runs on your computer with your models and files. The Saturn MCP gateway is a separate product, included in Pro and Enterprise, which are coming soon.

MCP Gateway for enterprise AI agents

Give your AI agents access to tools. Keep control of every action.

Saturn is a secure MCP gateway that sits between your AI agents and your MCP servers. Centralize access, enforce fine-grained permissions, require human approval for sensitive actions, and keep a complete audit trail — all from one console.

Get started Book a demo

Works with any Model Context Protocol (MCP) client. No changes to your existing MCP servers.

Compatible with

Claude
ChatGPT
Cursor
Gemini
LM Studio
Any MCP client

Connect once. Control precisely. Prove compliance.

Connect once

One MCP endpoint for every agent and every MCP server in your organization. Register a server, and its tools are available to govern immediately.

Control precisely

Decide which tools each agent can see and under which conditions it may call them — down to individual arguments like amount, environment, or recipient.

Prove compliance

Every request, policy decision, approval, and execution is recorded. Answer who did what, when, why it was allowed, and who approved it.

AI agents are powerful. Without a gateway, they are also unaccountable.

Most teams connect agents directly to MCP servers. It works in a demo. It does not hold up in production.

Without Saturn

Every agent holds its own API keys and OAuth tokens
Tool access is all-or-nothing per MCP server
No way to pause a sensitive action for human review
Logs are scattered across servers, if they exist at all
Revoking one agent means rotating credentials everywhere

With Saturn

Credentials stay in a post-quantum encrypted vault; agents never see them
Curated tool sets per agent, with argument-level policy rules
Built-in approval queue for sensitive calls, with full context
One audit trail covering every agent and every tool call
Pause an agent or reject a call instantly from the console

Everything you need to run AI agents in production.

Saturn covers the full lifecycle of a tool call: identity, access, policy, approval, execution, and audit.

01 / Unified MCP gateway

One endpoint for all your agents and tools

Standards-compliant MCP endpoint over HTTP — initialize, tools/list, and tools/call
OAuth 2.1 with PKCE for every agent connection, bound to a user or an organization
Register any MCP server by URL; Saturn discovers and catalogs its tools automatically
Managed upstream OAuth with dynamic client registration and automatic token refresh

02 / Fine-grained access control

Permissions down to the argument level

Tool sets: choose exactly which tools each agent can see in tools/list
Per-tool mode: mark any tool as Call (automatic) or Approval (human-confirmed)
Policy rules with conditions on arguments — amount > 1000, environment == "production"
Versioned policies with full change history, evaluated on every single call

policy    Payments  v3  production

rule      payments.transfer
when      amount > 1000 AND currency == "USD"
decision  approval_required

rule      payments.transfer
when      true
decision  allow

rule      db.drop_table
when      true
decision  deny

A policy as the engine evaluates it. Conditions are matched against the arguments of each call.

03 / Human-in-the-loop

Approvals built into the request path

Sensitive calls pause at the gateway until a reviewer approves or rejects them
Reviewers see the agent, the tool, the full arguments, and the rule that triggered the hold
Rejections are returned to the agent as explicit decisions, not timeouts
Pause any agent instantly; paused agents cannot start new calls

04 / Enterprise-grade security

Secrets never leave the gateway

Credential vault sealed with ML-KEM-1024 post-quantum key encapsulation and AES-256-GCM
APIs report whether a secret exists — never its value; every read is an audit event
Outbound protection: private, link-local, and cloud metadata addresses are blocked
Denied and pending calls are never forwarded upstream

05 / Audit and cost visibility

Complete visibility into what your agents do

Every step is an event: requested, evaluated, approval requested, granted or rejected, executed
Each event records the agent, the tool, the arguments, the policy version, and the reviewer
Wallet and x402 payment tracking shows what each tool call cost
Filter and export the trail for compliance and incident review

06 / Team and role management

Built for organizations, not just individuals

Preset roles — admin, operator, viewer — or custom permission sets
Per-resource access levels for agents, policies, approvals, and team
Scope editors to specific agents so teams only manage what they own
Map which MCP servers each employee is allowed to connect

Why a dedicated MCP gateway?

API gateways govern services. Saturn governs agents — a caller that reasons, chooses tools, and needs to be stopped sometimes.

Direct connection

Agent holds every credential
Every tool on the server is exposed
No approval step
No central audit trail

API gateway

Sees routes and methods, not tools
No conditions on tool arguments
No human approval workflow
Designed for services, not agents

Saturn

Credentials sealed in the gateway
Tool sets and argument-level policy
Built-in approval queue
One audit trail for every agent

One console for every agent in your organization.

Saturn Console gives security, platform, and operations teams a shared view of what agents can do and what they have done.

Overview

Active agents, pending approvals, recent denials, and spend at a glance.

Agents

Saturn MCP

Connect MCP servers, review discovered tools, and build tool sets.

Policy

Author and version rules; see which policy decided each call.

Intervention

Approve or reject pending calls with full context.

Audit

Search every event by agent, tool, decision, reviewer, or time.

Cost

Wallets, x402 payments, and per-call cost attribution.

Identity

Members, roles, custom permissions, and MCP server mapping per employee.

Plans for individuals and organizations

Start free today. Pro and Enterprise are coming soon.

Personal

Free

$0 · forever

For individuals running agents on their own computer.

Saturn App for macOS and Windows
Bring your own models and MCP servers
Built-in skills: files, code, web, MCP
Local credential broker
Open source, MIT licensed

Get started

Personal

Pro

Coming soon · per user / month

For professionals who need the gateway in front of their agents.

Everything in Free
Saturn MCP gateway access
Tool sets and per-tool approval mode
Policy engine with versioned rules
Approval queue and full audit trail

Organization

Enterprise

Coming Soon

For teams that need governance across the whole organization.

Everything in Pro
Members, roles, and custom permissions
Shared team agents with scoped editors
MCP server mapping per employee
Priority support and onboarding

Questions about Saturn

What is Saturn?: Saturn is a secure MCP gateway between AI agents and the tools they call. It centralizes access, enforces fine-grained permissions, requires a person to approve sensitive actions, and keeps a complete audit trail in one console.
What is an MCP gateway?: MCP (Model Context Protocol) is the standard AI agents use to call tools. An MCP gateway sits in front of your MCP servers. Agents talk only to Saturn, which exposes the tools you choose, evaluates every call against policy, and forwards only the calls that are allowed.
Which AI clients work with Saturn?: Any MCP client, including Claude, ChatGPT, Cursor, Gemini, LM Studio, and Saturn App. Every agent uses one Saturn MCP endpoint. Existing MCP servers do not need code changes.
How is Saturn different from an API gateway?: An API gateway governs services by route and HTTP method. Saturn governs agents. It understands tools and arguments, can hold a call for human approval, and records which policy and which reviewer allowed it.
Do agents see API keys or OAuth tokens?: No. Upstream credentials stay in Saturn’s vault, sealed with ML-KEM-1024 post-quantum key encapsulation and AES-256-GCM. Agents receive only the result of an authorized call. APIs report whether a secret exists, never the secret itself.
Is Saturn App free?: Yes. Saturn App is a free, open-source desktop agent for macOS and Windows, MIT licensed. It runs on your computer with your models and files. The Saturn MCP gateway is a separate product, included in Pro and Enterprise, which are coming soon.

Ready to bring your AI agents under control?

Connect an MCP server, build a tool set, write your first policy. See the full request path on your first approval.

Get started Book a demo