Stripe MCP server and Dome
Let agents read live Stripe data while their writes stay in a sandbox. Connect Stripe's MCP server to Dome twice: once with a live agent key, once with a sandbox key. Rules decide each call on its connection and its tool. A write aimed at live mode is refused at the Gateway.
Dome provides a tool gatewayand authorization for Stripe.
Live mode is read-only
A rule refuses writes on the live connection while reads go through.
Writes land in a sandbox
Agents create invoices and refunds against sandbox data, so nothing they write reaches a real customer.
Keys stay in Dome
Dome stores each Stripe agent key and injects it on the outbound call, so agents never hold one.
Stripe behind the Gatewayin three steps.
Add one connection per Stripe environment, sync the catalog, and apply the rule. Agents reach both through the Gateway's single MCP endpoint.
Add live and sandbox connections
Stripe fixes the mode by the key, so each environment gets its own connection. Use an agent API key with only the permissions the agents need.
shelldome tools add --name stripe-live \ --url https://mcp.stripe.com \ --auth-method api-key --credential-type shared \ --authorization "Bearer $STRIPE_LIVE_AGENT_KEY" \ --gateway finance-gateway dome tools add --name stripe-sandbox \ --url https://mcp.stripe.com \ --auth-method api-key --credential-type shared \ --authorization "Bearer $STRIPE_SANDBOX_AGENT_KEY" \ --gateway finance-gatewaySync the catalog
Dome calls Stripe's tools/list with the stored key. A key Stripe rejects fails here, before any agent traffic.
shelldome tools catalog sync stripe-liveApply the rule
Scope it to one agent while you try it, and simulate before you deploy.
shelldome rules apply stripe-live-read-only.cedar \ --agent invoice-assistant --name stripe-live-read-only
Read live,write in the sandbox.
The permit opens both Stripe connections to the agent. The forbid refuses stripe_api_write on the live connection, the tool behind every POST, PATCH, PUT, and DELETE.
cedar
permit (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "stripe-live" || resource.connection_name == "stripe-sandbox"};
forbid (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "stripe-live" && resource.tool_name == "stripe_api_write"};One call,two outcomes.
Switch the caller or the argument and watch the same call decide differently, with every decision landing in audit.
Connection
agent invoice-assistant · acting as a.nakamura
stripe-sandbox/stripe_api_write(Create an invoice)
- Agentinvoice-assistant is registered and active
- Callera.nakamura verified through Okta
- RuleSandbox writes are allowed
DecisionAllowed
Three agentson Stripe.
These come from our template library, and each one reads broadly and writes narrowly.
billing-support
Answer billing questions
Looks up the customer, their charges, and their invoices and drafts a reply, while refunds stay with a person.
revenue-reporter
Report on revenue
Pulls subscription and billing metrics and writes the weekly revenue summary.
invoice-assistant
Draft invoices
Reads the live customer record and builds the invoice in the sandbox for a person to check.
Bringing it alltogether.
Connecting Stripe to registered agents, models, and identity in Dome completes a governed agent application.
Acting for
Control point
GatewayEvery call decided and audited
RulesGuardsQuotas
Tools
StripeMCP server
Models
Common questionsabout Stripe.
More of whatDome works with.
Model
Claude Fable
Fable 5.1 from Anthropic and Amazon Bedrock in one failover pool, open to one group and capped by quota.
Provider
Anthropic
The Claude API behind the Model Broker: the key held in Dome, every call authorized, metered and audited.
Identity
Microsoft Entra ID
Entra access tokens verified on every agent call, so rules read app roles and audit names the person.
Runtime
LangGraph
LangGraph agents with their model calls on the Model Broker and their MCP tools on the Tool Gateway.
Client
Claude Code
Claude Code on a Dome Gateway with per-developer sign-in, rules on every tool call, and audit by name.
Agent service
TinyFish
TinyFish's web agents behind the Tool Gateway, with rules that decide each run on the site it targets.
Talk with ourFDE team.
Or start free, and register your first agent in minutes. Connect it through a gateway, give it rules, and see every call it makes in one audit trail.
Talk with our FDE team
Our engineers design, build and run governed agents for the part of your business you want to change. You see what they deliver.
MCP servers for AI agents: per-argument rules on every call
Put any remote MCP server behind the Tool Gateway. Agents reach it through one governed endpoint, and rules decide each call on its arguments.