SatGate Agent Control Plane
Treat every model like a capable insider that can make mistakes or be tricked. SatGate gives each agent its own limits, checks each request against them before it reaches your services, and lets you stop an agent and every helper it started from one place.
Your agents do the work. You keep the authority.
Example: one agent and one helper
- Research agent: 5,000 credits, 3 tools, ends Friday
- Helper: 1,000 credits carved from those 5,000, and 1 of those tools
- The helper calls a tool it wasn't given: refused, with a signed receipt
- You revoke the research agent: the helper's next request is refused too
An example of limits you set in SatGate. Your numbers and tools will differ.
Limits the model can't change.
Budgets, allowed tools and end dates live in SatGate, not in a prompt or a provider's settings. SatGate checks each request against them before it reaches the service. In Control, a request outside those limits is refused. Observe records it and does not block on budget.
Handoffs that only narrow.
An agent can give a helper part of what it has: the same tools or fewer, and a budget carved out of what it has left. It can never pass on more.
One off switch for the whole chain.
Revoke an agent and its next request is refused, along with every helper it handed work to. If a revoke can't be confirmed, the dashboard says so instead of showing it as done.
Evidence the model didn't write.
SatGate signs a receipt when it refuses an agent over a budget, a missing payment or a tool permission, and when an agent spends from a budget in Control. The agent never writes or edits its own record. A SatGate Cloud receipt can be checked with the open-source verifier against our published keys.
Why not just give the agent an API key?
A key in an agent's settings works for anyone who copies it, has no budget, and keeps working until someone rotates it. With SatGate, the key to your service stays in SatGate. The agent gets its own token, with its own limits and an end date, and stopping it doesn't mean changing the key your other agents and apps use.
Who it's for
Platform teams: One place to connect agents to your APIs and MCP tools, with limits per agent instead of shared keys.
Security teams: Signed receipts for budget, payment and tool refusals, the chain of handoffs in the dashboard, and an off switch that reaches every helper.
Finance teams: Budgets checked before the spend in Control, and spending per agent on the Usage page.
Questions
What is an agent control plane?
It's the one place where you decide what each AI agent can reach, how much it can spend, who it can hand work to and when its access ends, and where you stop it. SatGate checks those limits on each request, before the request reaches your services.
How is an agent control plane different from an API gateway?
An API gateway protects one service. SatGate follows the agent: one budget across the services it calls, the helpers it hands work to, and an off switch that reaches all of them.
Why shouldn't AI agents hold API keys?
A key works for anyone who holds it, has no budget and lasts until someone rotates it. With SatGate, each agent gets its own token with its own limits and end date, and you can stop one agent without changing the keys anything else uses.
How does SatGate control handoffs between agents?
An agent can give a helper only part of what it has: the same tools or fewer, and a budget carved out of what it has left. Revoking an agent refuses its helpers' next requests too.
Related pages
Rules and Receipts
How SatGate checks each request against your rules, and when it signs a receipt.
MCP governance
Budgets, allowed tools and an off switch for your agents' tool calls.
Agent API governance
Give each agent its own limited token instead of a shared API key.
Agent capability tokens
What an agent's token carries: the routes it can use, the budget it spends from and its end date.
Revocable agent credentials
Stop one agent without changing the keys anything else uses.
Economic firewall
Budgets for your agents, and payment from external agents, on the routes you choose.
AI agent cost control
Set budgets for your agents and their helpers, and see what each one spends.
Economic firewall readiness grader
Score how ready your agent setup is, from budgets and revoke to receipts and MCP.
SatGate for Hermes Agent
Put limits on local Hermes agents, with SatGate checking each request before it goes through.
Give every agent its own limits, and keep the off switch.
Start in Observe to see what your agents do, then turn on Control when you're ready.