Any agent. One platform.
The Agent Operations Platform
Dome is the composable agent platform for consistent agentic operations, built for AI Platform Teams. Agent Registry, MCP Gateway and Model Broker, assembled from one set of resources, whatever the framework, provider or cloud.
Platform resources
Build your Agent Factory with Dome
Dome provides a foundation for your teams to build the shared services they need to power your AI transformation. If you're building an agent factory, Dome provides a flexible blueprint for operations.
Connect
Govern
Operate
Connect · Resource
Agents
The registered identity of an agent.
Name, owner, declared capabilities and a lifecycle: provisioned, active, suspended, revoked. The key issued at registration is the root of trust for every call. Revoke the agent and every call stops. Where it fits →
Agents built on
$ dome agents register --name support-triage$ dome agents create-key support-triage --name prodConnect · Resource
Callers
The person an agent is acting for.
Attach your identity provider, and callers are verified against it over OIDC (or by HMAC for services) before they become an identity. A rule can then name the caller as well as the agent, and the audit record names both. An agent can be required to always act for someone. Where it fits →
Attach your identity provider
$ dome agents update support-triage \ --actas-required \ --actas-method oidcConnect · Resource
Tools
A connection to something an agent can call.
An MCP server, a REST API described by OpenAPI, or an internal service. Tools surface in a catalog that policy filters per agent, so an agent sees only the tools it may call. Where it fits →
$ dome tool add --name zendesk \ --url https://mcp.zendesk.example.com \ --protocol streamable-http \ --auth-method oauthConnect · Resource
Models
A connection to a model provider.
Anthropic, OpenAI, Bedrock, Vertex, Azure OpenAI, or any OpenAI-compatible endpoint, including your own. The provider credential lives with the connection and is injected at egress. Where it fits →
Route to
$ dome model add anthropic-prod \ --provider anthropic \ --model claude-sonnet-5Connect · Resource
Pools
A named group of models with a routing strategy.
Priority and weight choose the member that serves a request, and failover falls through on rate limits or outages. Each candidate is authorized in its own right, so failover can't route around policy. Where it fits →
Pool model families like
$ dome model pool create prod-models \ --routing-strategy priority_weighted$ dome model pool member add prod-models anthropic-prod --priority 0Connect · Resource
Gateways
The named access surface you create and grant.
A gateway composes tools and model pools into one endpoint an agent is granted as a unit. Gateways are logical, so make as many as your estate needs. A resource in no gateway is unreachable. Where it fits →
$ dome gateway create support-prod$ dome gateway tool-sources add support-prod zendesk$ dome gateway model-pools add support-prod prod-models$ dome gateway access grant support-prod support-triageGovern · Resource
Rules
Policy evaluated at the call boundary.
A rule permits or forbids a call before it runs, on the agent, the caller, the tool and its arguments. Written at org, tenant, workspace or agent scope. Forbid wins, and the default is closed. Simulate before you ship. Where it fits →
permit ( principal == Dome::Agent::"support-triage", action == Dome::Action::"mcp:call", resource == Dome::MCPTool::"zendesk/update_ticket") when { principal has act_as && resource.arguments.requester == principal.act_as.email};Govern · Resource
Guards
Inspection and filtering of content in flight.
Classify, redact or constrain a request or response before it passes, so a permitted call can't return data the caller shouldn't have. Versioned with rollback, assigned per connection and direction. Where it fits →
Filter calls to services like
$ dome guards filters create pii-redact \ --config-from pii.json$ dome tool guards filters set zendesk \ --direction response --filters pii-redactGovern · Resource
Quotas
Spend and usage ceilings bound to an identity.
Bound to an agent, workspace or tenant, not to an API key, so attribution comes free. A denial at the limit is recorded like any other decision. Where it fits →
$ dome model quota set \ --subject agent \ --agent support-triage \ --limit 2000000 \ --window 24hOperate · Resource
Audits
The event trail, after the fact and live.
One vocabulary for every call: agent, caller chain, target, arguments, rule version, outcome. Reconstructing what an agent did is one query. The same trail streams live, and exports continuously or on a schedule to the SIEM and observability tools you already run. Where it fits →
Export the audit trail to
$ dome audit query$ dome audit chain act_9f2k1p$ dome audit streamOperate · Resource
Webhooks
Signed events pushed to your endpoints.
Subscribe to the event types you care about and receive them signed, with retries and replay. A denial reaches an on-call channel or a runbook without anyone polling. Where it fits →
$ dome webhooks subscriptions create siem \ --url https://siem.example.com/dome \ --event-type tool.call --filter 'result=denied'Operate · Resource
Integrations
The systems Dome composes with.
SIEM, observability, incident and automation tools. Dome runs alongside what you operate: the audit trail and events land where your teams already look. Where it fits →
Operate · Resource
Assistants
AI assistance over the platform itself.
The Rules assistant drafts, validates and refines rules from the dashboard. What it produces goes through the same validation and simulation as a hand-written rule: it drafts, it doesn't deploy. Where it fits →
$ dome rules validate ./rules/support-triage.cedar$ dome rules simulate --agent support-triage \ --action mcp:call --resource zendesk/delete_ticketHow it works
How Dome governs agents
Three steps, in the order a platform team takes them. Each diagram lights what its step sets up: solid boxes are Dome resources, dashed ones are yours, and every box opens its docs.
Step 1
Connect
Give agents governed access. Every agent is known, and every tool and model is reached through a governed path. Register agents, verify the people they act for, and put tools and models behind gateways, in that order.
Step 1: Register the agent
Step 2
Govern
Enforce who can call what. Every call is evaluated against policy before it executes. Rules, guards and quotas apply at every hop, the same way for every agent.
Step 3
Operate
Track, audit and act on every call. Every decision is recorded and visible in the tools you already run. One record for every call, queried in Dome or sent where your teams already look.
Step 1: Query the record
Next steps
Talk with our FDE team
Our forward deployed engineers work with your platform team to get your agents into production and under control: the first one governed on your own systems, and a pattern your teams can repeat for every agent after it.
No card required to start. Register your first agent in minutes.
Application patterns
Golden paths for the agent you are building, each wired from the same resources.
See the patterns