AWS has announced a significant enhancement to Amazon Bedrock AgentCore Runtime with the introduction of Interactive Shells, giving developers direct terminal access to running AI agent sessions. The new capability is available through the InvokeAgentRuntimeCommandShell API and enables a persistent terminal experience inside the isolated environment where AI agents execute.
This update represents an important step forward for developers building and operating coding agents, providing a more familiar and powerful way to interact with AI-powered development environments.
What Is New?
Previously, developers could execute commands within an agent runtime using the InvokeAgentRuntimeCommand API, which was designed for one-time command execution.
With the new Interactive Shell capability, developers can now connect to a persistent terminal session running inside the agent's microVM. Instead of executing individual commands separately, users can interact with the environment as if they were working on a local machine.
The terminal is PTY-backed and supports familiar terminal features including:
Full color output
Tab completion
Ctrl+C interruption support
Terminal window resizing
Automatic reconnection during temporary network interruptions
Persistent command history
This creates a much more natural experience for debugging, development, and operational tasks.
Why This Matters
As AI coding assistants become increasingly capable, developers often need visibility into what those agents are doing behind the scenes.
Interactive Shells allow developers to:
Connect directly to the runtime environment hosting the agent
Inspect generated files
Review application state
Execute ad-hoc commands
Troubleshoot issues in real time
Monitor agent activity during complex tasks
Instead of treating the AI agent as a black box, developers can now actively inspect and interact with the environment while the agent is working.
Built for Modern Coding Agents
The feature is especially useful for organizations running coding agents such as:
Claude Code
OpenAI Codex
Amazon Kiro
Custom AI development agents
Developers can authenticate into the runtime environment and interact with the same workspace being used by the agent, making debugging and collaboration significantly easier.
Persistent Sessions and Reconnection Support
One of the most valuable aspects of the new feature is session persistence.
Environment variables, working directories, and command history remain available throughout the session. This means developers can execute multiple commands over time without losing context.
Each shell session is identified using:
Runtime Session ID
Shell ID
If a connection is interrupted, developers can reconnect using these identifiers and return to the exact same terminal session. Short network disruptions are handled automatically, while longer disconnects can be resumed manually.
Multiple Concurrent Shells
AgentCore Runtime now supports up to 10 concurrent shell sessions per agent runtime.
This allows developers to:
Open multiple terminal windows simultaneously
Monitor different tasks in parallel
Compare outputs across environments
Work with multiple agent branches at the same time
For teams managing complex AI workflows, this provides greater flexibility and operational visibility.
What This Means for Developers
The addition of Interactive Shells transforms AgentCore Runtime from a command execution platform into a more complete development environment for AI agents.
Developers gain direct access to the runtime environment while maintaining the isolation and security benefits provided by AgentCore's microVM architecture. The result is a workflow that feels much closer to working with a traditional development machine while leveraging the power of autonomous AI agents.
Final Thoughts
The introduction of Interactive Shells is a practical and developer-focused enhancement to Amazon Bedrock AgentCore Runtime. By providing persistent terminal access, session recovery, and support for multiple concurrent shells, AWS is making it easier to build, monitor, and debug AI-powered coding agents at scale.
As organizations continue adopting AI-assisted software development, features like Interactive Shells help bridge the gap between autonomous agents and traditional developer workflows, giving teams greater control, transparency, and productivity.