Crafting Introduces Express for Local AI Agent Orchestration

Overview

Crafting has introduced Express, a new local edition of its AI agent orchestration platform, designed to allow teams to evaluate multi-agent coordination within a single Docker container. The release offers developers a streamlined way to deploy, test, and inspect agent handoffs without the complexity of enterprise-grade infrastructure.

According to the company, Express serves as a lightweight entry point into the Crafting ecosystem, enabling users to launch specialist agents, manage parallel workflows, and utilize a integrated web interface and command-line tool. It is important to note that this edition is strictly intended for evaluation and trial purposes; production deployments require the separate Crafting Enterprise tier.

The platform addresses a growing need for structured agent collaboration, moving beyond simple chatbot interactions to complex, multi-step task execution. By containerizing the orchestration layer, Crafting aims to lower the barrier to entry for engineering teams looking to integrate AI agents into their existing development pipelines.

Key Features and Capabilities

Crafting Express brings core orchestration functionalities to a local environment, focusing on three primary areas: specialized delegation, model flexibility, and collaborative workspaces.

Delegating Work to Specialist Agents

A central pillar of the Crafting platform is the ability to split complex tasks among dedicated sub-agents. In this architecture, a lead agent acts as the orchestrator, breaking down a request and assigning specific components to coding, testing, or security specialists.

These sub-agents operate in parallel, each maintaining its own conversation context. Once they complete their assigned work, results are returned to the lead agent, which then synthesizes the findings and issues follow-up instructions as necessary. This model mimics a human team structure, allowing for more robust and specialized problem-solving than a single monolithic agent could achieve.

Model-Agnostic Configuration

Teams can choose their preferred AI models without having to replace their underlying execution harness. Crafting’s native support includes OpenRouter, compatible self-hosted endpoints, AWS Bedrock, and Google Cloud Vertex AI.

This flexibility allows organizations to experiment with different language models to find the optimal balance between cost, speed, and capability. Centralized configuration management ensures that agent definitions and execution environments remain stable even as teams swap out the foundational models powering them.

Collaborative Workspaces

Unlike many isolated AI tools, Express supports shared environments where agents and human engineers can work simultaneously on the same codebase. Agents can be granted dedicated environments or provided access to existing workspaces, fostering a collaborative workflow.

Engineers can use standard development tools alongside their AI agents, validating changes in real-time. This setup is particularly useful for iterative development cycles, where an agent might propose a fix while a developer reviews the code and provides immediate feedback or adjustments.

Security and Infrastructure Control

Crafting emphasizes granular control over agent capabilities, ensuring that automation does not come at the cost of security. The platform allows administrators to scope tools and access levels to match each agent’s specific role.

Sandbox and Access Controls

The integration combines sandbox access controls, network isolation, and managed credentials to constrain what agents can reach. When agents coordinate across different systems, they operate within strictly defined boundaries that prevent unauthorized data access or unintended side effects.

Credentials are scoped and injected per agent, meaning administrative secrets are never exposed directly to the AI models. This approach aligns with security best practices for autonomous software, reducing the attack surface while enabling sophisticated cross-system operations.

Production-Like Testing Environments

Agents can connect to real services and dependencies within production-like environments. This capability allows them to run actual tests, inspect failures, and iterate on fixes rather than relying on simulated responses.

Both agents and engineers can validate changes in parallel using isolated slices of the application stack. This feature is critical for teams looking to deploy AI-driven maintenance or enhancement tools, as it provides a safe sandbox for experimentation before changes impact live systems.

Getting Started with Express

Launching the Express edition is designed to be frictionless, utilizing a single command after installing the Crafting CLI. The setup process guides users through signing up, creating Docker volumes for persisted states, and initializing the container.

curl -sSL https://sandboxes.cloud/sh/install-cli | bash
cs express start

Upon initiation, the CLI provides clear instructions on accessing the web UI and managing agents via the command line. For detailed documentation, users are directed to the project’s official repository, which includes troubleshooting guides and advanced configuration options.

From Evaluation to Enterprise

While Express covers the fundamentals of agent orchestration, Crafting acknowledges that larger organizations will eventually outgrow a single-container solution. The company highlights several limitations of the Express edition, including the lack of multi-cloud support and enterprise-grade security audits.

Teams hitting these limits are encouraged to upgrade to Crafting Enterprise, which offers:

• Kubernetes Integration: Agents execute against real cluster infrastructure rather than local containers.
• Multi-Cloud and Multi-Region Support: Automated failover capabilities ensure high availability without vendor lock-in.
• Enhanced Security: SOC 2 Type II certified infrastructure with admin-managed credential injection and zero exposure of secrets to agents.

Crafting promises a deployment timeline of two weeks or less for Enterprise customers, aiming to bridge the gap between local experimentation and production-scale AI automation.

Industry Context

The introduction of Express comes amid a broader industry shift toward agentic AI—systems that can plan, execute, and refine tasks autonomously. As organizations move from prototyping single AI interactions to deploying complex agent swarms, tools that simplify orchestration and security become essential.

Competitors in the space often focus heavily on either the user interface or the underlying model routing. Crafting differentiates itself by providing a unified platform that handles both the workflow logic and the infrastructure constraints, ensuring that agents remain secure and auditable throughout their lifecycle.

For developers and engineering leads, Express offers a low-risk opportunity to assess whether agentic workflows can enhance their current processes. By removing the initial overhead of enterprise procurement and complex infrastructure setup, Crafting is positioning itself as a pragmatic choice for teams ready to explore the next generation of software development tools.