Devin MCP and Dome
Devin works in your systems, and you decide what it touches. Devin writes code on its own and reaches your tools through custom MCP servers. Point it at a Dome Gateway and it works there as a registered agent: pull requests in your organization, no merges, and every call on the record.
Dome provides a governed gatewayfor Devin.
Devin signs in as Devin
Devin carries an agent key of its own, not an engineer's token, and revoking it ends Devin's tool access.
Tools scoped by rule
Rules name the tools Devin may call and read their arguments. Pull requests open in one organization, and merges stay with people.
One trail with your agents
Devin's calls sit in the same audit trail as your own agents', allowed and refused alike.
Devin on a Gatewayin four steps.
Register Devin, give it a key, apply the rules, and add the Gateway to Devin as a custom MCP server. Your tools stay behind the Gateway the whole time.
Register Devin
Give it the GitHub tools it may call and the Gateway in front of them.
shelldome agents register --name devin \ --tool github/search_code \ --tool github/get_file_contents \ --tool github/create_pull_request \ --gateway eng-gatewayCreate its key
The response holds the key and the MCP URL. Paste both into Devin and keep no other copy.
shelldome agents create-key devin --name devin-mcp --gateway eng-gatewayApply the rules
Scope them to the devin agent. Simulate a merge to see it refused, then deploy.
shelldome rules apply devin-github.cedar \ --agent devin --name devin-githubAdd the Gateway to Devin
In Devin, open Customize, then MCPs, and add a custom MCP. Choose HTTP, paste the MCP URL, and set the Authorization header to Bearer and the key.
Read code, open pullrequests in one org.
The permit opens three GitHub tools to Devin. The first forbid refuses every other GitHub tool, merges included, and the second refuses a pull request outside the acme organization.
cedar
permit (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "github" && ["search_code", "get_file_contents", "create_pull_request"].contains(resource.tool_name)};
forbid (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "github" && !["search_code", "get_file_contents", "create_pull_request"].contains(resource.tool_name)};
forbid (principal, action == Dome::Action::"mcp:call", resource is Dome::MCPTool)when { resource.connection_name == "github" && resource.tool_name == "create_pull_request" && !(resource has arguments && resource.arguments has owner && resource.arguments.owner == "acme")};One call,two outcomes.
Switch the caller or the argument and watch the same call decide differently, with every decision landing in audit.
Organization
agent devin · scheduled
github/create_pull_request(owner: "acme", repo: "web", head: "devin/fix-retry", base: "main")
- Agentdevin is registered and its key is active
- Gatewaydevin may reach eng-gateway
- RulePull requests are open in acme
DecisionAllowed
Bringing it alltogether.
Connecting Devin to registered tools, models, and identity in Dome completes a governed agent application.
Acting for
Agent
Control point
GatewayEvery call decided and audited
RulesGuardsQuotas
Tools
DevinAgent service
GitHubMCP serverModels
Common questionsabout Devin.
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.
MCP server
Atlassian
The Atlassian Rovo MCP server behind the Tool Gateway, with rules that decide each Jira and Confluence call on its tool and site.
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
Codex
Codex on a Dome Gateway with per-developer sign-in, rules on every tool call, and audit by name.
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.
Agent services: web, research, payments, voice and sandboxes
Web actions, research, payments, phone calls and code execution are where agents reach past your walls. Put each service behind the Gateway, and every call is authorized, metered and audited.