Back to the blog

Kiro adopts ACP: agent interoperability reaches the specialized IDE

Kiro announced Agent Client Protocol (ACP) support in its CLI. The change lets you use the agent's capabilities from compatible IDEs and tools, instead of tying them to a single editor. For technical teams, the relevant change is architectural: a common protocol reduces ad hoc integrations and lets you apply the agent where specialized context already exists.

From one-to-one integrations to a common contract

Before a shared protocol, every combination of agent and editor requires a specific integration. ACP proposes standardizing communication between clients —editors or tools— and coding agents. Kiro compares this role with the impact the Language Server Protocol had on language capabilities in IDEs.

The consequence for an organization is less dependence on a single interface. A team can keep its specialized environment for databases, data science or mobile development and reach the agent through the same contract, instead of giving up working context to adopt an AI tool.

The important use case is specialized context

The announcement illustrates the integration with DataGrip: the tool provides schema visualization, execution plans, autocompletion and connection management; the agent can help with queries, migrations, test data or explaining complex conditions. Neither layer replaces the other.

This pattern applies to other domains. The specialized tool keeps the area's own model, telemetry or controls. The agent brings reasoning and automation on top of that context. The value isn't in putting chat everywhere, but in shrinking the distance between intention, context and verifiable action.

Interoperability demands execution controls

An agent that can run commands from a tool connected to real data widens the risk surface. The protocol doesn't remove the need for identity, authorization, secure secrets, isolated environments and review of actions with impact.

To adopt this capability professionally, it pays to define which environments the agent can touch, which actions require approval, how activity is logged and who is responsible for reviewing a change. Interoperability speeds up the flow; governance keeps it from speeding up mistakes.

Reference architectureInteroperability between a specialized tool and an agent

A compatible IDE or tool implements ACP and talks to Kiro CLI. The agent receives context from the tool, proposes or runs authorized actions and returns results to the specialized environment.

Operational impact: ACP lets you place agents where the working context is. The next challenge for teams is combining that flexibility with permissions, traceability and execution controls proportionate to the risk.

Keep reading

More in AI and agents Reading path: Engineering with AI and agents