Skip to content
Platform

Connect identity, policy and action.

Discover what is running, put one check on the execution path, and keep evidence for every decision.

Explore the platform Plan with Abi

One connected platform

Five stages, one policy. Start anywhere and add the rest as you need it.

  1. Connect
    Gateway and SDKs on the execution path.
  2. Identity
    Who is calling, and for whom.
  3. Govern
    One policy, tested before release.
  4. Observe
    Discovery, change review, coverage.
  5. Respond
    Investigate and act on what is found.
Connect

Two enforcement points. One policy.

The gateway protects requests travelling to services. The SDK protects actions inside the agent application. Both use the same AgenticDome policy and runtime services, so the choice is where the check sits, not which rules apply.

Action Gateway A sidecar gateway for MCP, HTTP APIs, GraphQL and A2A. It inspects requests and, where supported, responses, and uses signed policy from the control plane. No SDK hooks are needed for calls to services. MCP tool-call arguments are checked against published input schemas before forwarding. Compiled validators are cached locally; optional schema pins let security teams approve sensitive tool contracts.
SDKs and native plugins Hooks and tool wrappers inside the application provide checks for connected local functions, framework context and execution lineage. Docker Agent can attach its mandatory native tool guard through deployment configuration to check local shell, file and tool proposals without application-code changes.
Where should the check sit?
If your agentUse
Calls MCP servers, APIs, GraphQL or other agentsGateway
Takes consequential actions inside application codeSDK or runtime plugin
Runs Docker Agent shell, file or local toolsDocker mandatory native guard + assigned sidecar
Does bothGateway and SDK
Might go around the checkAdd Runtime Grid discovery and governed egress
See the layers as an interactive graphic on the home page
Identity

Reuse the identity you already have.

Use tokens from your identity provider, such as Entra ID or Okta, or identities issued by AgenticDome. Verified caller identity and available delegated authority inform connected gateway and supported SDK decisions. Deployments, provider trust, credential rotation and revocation are managed, and history is kept.

Verified Action Chain

From discovery to outcome, six steps.

  1. 1 · Discover
    Find the agent and its tools.
  2. 2 · Attest
    Verify who is calling.
  3. 3 · Authorise
    Decide against policy.
  4. 4 · Admit
    Check exact-request authority before the side effect.
  5. 5 · Execute
    Forward or run the admitted request once.
  6. 6 · Verify
    Record the outcome as evidence.

Outcome evidence is kept distinct by source: reported by the SDK, observed at the gateway, or destination-attested through a configured destination adapter that signs a receipt of the business result. Enforcement applies only to execution paths connected through supported integrations. See the workflow control points →

Govern

One action policy, released with care.

Shared action policyThe same rules at the gateway and in the SDK.
Controlled releasesTest policy changes and review them before they reach production.
Human approvalHold eligible high-consequence actions for governed human approval.
Runtime Intelligence Grid

Three kinds of evidence, one investigation.

A network connection alone does not show intent or business outcome, so the Grid correlates three layers.

ApplicationDecisions from SDKs and integrations.
WorkloadProcess and connection evidence for selected Linux workloads.
NetworkDestination and enforcement evidence from egress controls.
Observe

What exists, what is used, who called it, and what changed.

Connect a source and AgenticDome finds agents, MCP servers and models. Each record separates what is configured from what is actually used, and keeps the evidence that found it. Sample data below.

NameFound byExistsIn useWho called itExecution path
refund-agentSDK, gateway, Runtime GridConfigured4 minutes agoWorkload and 2 usersProtected, verified
analytics-mcpMCP inspection, registry card4 tools, 2 newYesterdaysales-research-botNot governed
invoice-extractorRuntime GridProcess seenProvider traffic seenWorkload onlySuspected bypass
Criticalanalytics-mcp added a delete tool
Before: 2 read-only tools. Now: 4 tools, including delete_records. 1 agent can call it, and it is not governed.
Reviewlegacy-report-writer not seen in 90 days
Stale record. Review, defer or reopen.
Tracked changesModel, destination, permission and protection changes
An agent switches model provider, an identity gains a permission, an A2A agent advertises a new skill, or a protection connection is removed. Each shows before, now, affected assets and a suggested action.
Who can call itsales-research-botSalesforce agent, acts for sales staff
analytics-mcp exposeslist_tables · Readrun_report · Readexport_table · Read, bulkdelete_records · Delete
What it can reachSales data storeCustomer and pipeline records

Finding something never grants it access or trust. Importing telemetry gives visibility. Active protection starts only when the execution path is connected.

From discovered to protected

Discover, connect protection, verify a real action, monitor.

DiscoverAvailable
Connect a source. AgenticDome finds agents, MCP servers and models, and records how each was found.
Respond

Investigate and act with scope.

Risks and investigationsFollow an action from discovery to outcome.
Action reviewsSee what was allowed, blocked or sanitised, and why.
Scoped responseUse supported, scoped response controls for the affected agent or tool.
Meet Abi

Your deployment concierge.

Tell Abi what you need your agents to do. Abi compares gateway, SDK, Docker native guard, identity and runtime options, explains the trade-off, and produces a deployment blueprint for your team to review. Ask it about MCP schema checks and how to verify each boundary. AgenticDome Copilot then guides setup and verification.

See how Abi and Copilot work together
Sample I need our refund agent to issue refunds, but only up to an approved amount. Best fit: gateway for the payment service, SDK on the refund function. Same policy at both. Want the blueprint?

Deployment models

Managed regionalUse an available published region assigned to your tenant.
DedicatedContracted isolated capacity for your tenant.
Customer-controlledContracted deployment within your approved environment.

Deployment FAQ

What does "protected" mean?
An asset is protected when its execution path is connected through a supported integration and a real action has been verified. Coverage applies only to connected paths.
Does importing telemetry protect my agents?
No. Imported telemetry gives visibility. Active protection needs the gateway or an SDK on the execution path.
What happens if the gateway cannot reach the control plane?
The gateway evaluates locally using its accepted signed policy and security state. It admits an action only while the required policy, identity and execution authority remain valid. Test readiness, expiry and outage behaviour for your deployment.
What is the performance impact?
Decision cost depends on placement, connection reuse, payload size and enabled controls. Local sidecar placement reduces network distance; measure allowed and blocked requests with representative traffic in your selected deployment.
Where is data stored?
Use your assigned published region. Dedicated and customer-controlled deployments have agreed residency, retention and support-access settings. Review content inspection and telemetry retention for the selected deployment.
How is usage billed?
The published plan meters distinct billable verified-action IDs. Duplicate evidence carrying the same ID is de-duplicated. Discovery records and imported telemetry describe visibility; optional services and deployment capacity follow the published plan or contract.

Start with what you can see.

Get started Book a deployment review