Basemachina Launches AI-Agent Skills, CLI, and API for Secure Internal System Operations

Basemachina (Headquartered in Chuo-ku, Tokyo; President: Seiji Takahashi), a provider of business infrastructure for internal management dashboards and product operations, has begun offering 'Agent Skills,' a 'CLI,' and a 'Public API' that allow AI agents to securely define and call operations on internal systems (e.g., internal databases, APIs, file storage).

This shift moves Basemachina from a primarily low-code UX for building internal admin panels to providing an AI-agent-first editing experience.

With this update, 'actions'—the calling configurations for internal tools (SQL or API calls)—can be managed as TypeScript code, allowing AI agents to create action definitions. The generated configurations can then be reviewed by humans via GitHub Pull Requests before implementation.

Furthermore, registered actions can be executed via the public API from external systems, CI/CD jobs, in-house scripts, or AI agents. Permission settings and approval workflows can be applied to all operations, making it an effective gateway for AI agents.

## Key Features

### Code-Based Configuration Management
Action settings can be managed as TypeScript code in a Git repository. The 'bm' CLI tool enables automated diff detection, dry-run previews before implementation, staged rollouts from development to production environments, and early error detection via type-checking.

### Agent Skills
These are skills designed for AI agents to handle Basemachina configuration creation. Provided under the name 'bm-code-management,' it supports Claude Code, Codex, Cursor, GitHub Copilot CLI, Gemini CLI, and Antigravity.

### Public API
This REST API allows for external interaction with environment lists, action lists, action details, and action execution. It can be used for routine CI/CD tasks, integration with in-house scripts, and tool execution by AI agents.

### Service Accounts
AI agents and external systems can be treated as execution entities independent of human users. They allow for individual management of permissions and audit targets, and support authentication using short-lived tokens linked to OIDC ID Tokens.

## Use Cases

- Handling requests for internal tool improvements: AI agents can create configuration proposals based on requests like 'create an action to search for users by email,' allowing human managers to simply review the changes.
- Routine data collection and operational tasks: Automate daily reporting or notification tasks via the public API from CI/CD jobs.
- Delegated research and information gathering: AI agents can perform research across multiple internal systems by calling pre-defined actions.

## Towards AI-Native Organizational Governance

When delegating internal system and data operations to AI agents, challenges such as excessive privilege granting, human error, and opaque execution history arise.

Basemachina addresses this by abstracting the scope of operations through actions, managing permissions per service account, enabling human review settings, and providing audit logs, allowing administrators to maintain control over what AI agents can execute. Basemachina aims to realize safe AI operational management where humans and organizations remain in charge of what is approved and executed, even as AI takes on the workload.