As AI agents move beyond answering questions to modifying code, processing business data, and coordinating with one another, the need for clear boundaries becomes critical. Natural-language instructions alone often leave contradictions undetected until runtime. Netsujo, a Kyoto-based technology company, addresses this challenge with Agent Role Contracts, an open-source tool that lets developers check role and scope declarations before any agent begins work.
Agent Role Contracts v0.5.0 is now available under the MIT License on GitHub and npm. It provides deterministic, offline checks that require no API key, AI model, or running agent. Developers can run a short demonstration to see how the tool flags a forbidden file-write request or an invalid review assignment, helping teams catch configuration problems early and build more trustworthy multi-agent workflows.
What is Agent Role Contracts?
Agent Role Contracts turns role and workflow declarations into structured data that can be inspected before execution. Instead of relying on natural-language prompts that may contain hidden ambiguities, developers define task scopes, agent responsibilities, review assignments, and handoff rules in a formal format. The tool then checks those declarations for contradictions and reports specific diagnostics such as TASK_WRITE_SCOPE_OUTSIDE_AUTHORITY.
Also worth a look: Incorpify Review: AI-Powered Company Formation & Management and IHL Hiring Demand Index Review: Weekly Hiring Pulse for Retail.
The checker is built as a Node.js package (requires version 22.5 or newer) and offers both a command-line interface and a JavaScript API. It is designed to integrate into existing development pipelines and CI processes. The initial stable release includes checks for role authority, task scopes, review separation, and handoffs, along with optional filesystem-write declaration adapters.
What stands out
Deterministic conflict detection
Agent Role Contracts performs checks deterministically and offline. There is no reliance on probabilistic AI models; the tool evaluates declared rules against requested actions in a predictable manner. This gives developers a reliable safety net before handing responsibilities to agents that can modify files or use external tools.
No API key or AI model required
Unlike many modern AI tools, Agent Role Contracts runs entirely with open-source code and standard Node.js. There is no need to register for an API key, configure a language model, or maintain an internet connection after the initial download. This makes it suitable for air-gapped environments and teams that prioritize data sovereignty.
Concrete diagnostics for common conflicts
The tool outputs specific diagnostics rather than vague warnings. For example, if a development agent is allowed to write only within a src/** directory but is handed a task to modify secrets/production.txt, the checker returns TASK_WRITE_SCOPE_OUTSIDE_AUTHORITY. It can also flag cases where the same declared role is responsible for both implementing a change and reviewing it, helping teams enforce separation of duties.
Who it’s for
Agent Role Contracts is aimed at development teams building or operating multi-agent systems where AI agents perform tasks such as code changes, data processing, or business support. It is especially valuable for teams that need to maintain audit trails of agent permissions and want to catch configuration errors before they lead to unintended behavior.
The tool’s open-source nature and permissive license make it accessible to startups, enterprises, and individual developers alike. It fits naturally into CI/CD pipelines, automated testing frameworks, and any workflow where role declarations are version-controlled alongside code.
Pricing and availability
Agent Role Contracts is completely free and open-source under the MIT License. There are no paid tiers, subscription fees, or usage limits. The source code is hosted on GitHub at github.com/suirindo/agent-role-contracts, and the npm package is available as @netsujo/agent-role-contracts. Developers can clone the repository, run the included three-minute demonstration, and start integrating the tool immediately.
Netsujo encourages technical feedback, reproducible failure cases, and community contributions. The project is actively maintained and has a clear roadmap for future checks and adapters.
What makes it worth considering
Agent Role Contracts addresses a fundamental gap in AI-agent orchestration: the lack of a lightweight, deterministic way to verify role boundaries before execution. As agents gain capabilities to act on external systems, teams need more than prompt engineering to ensure safety. This tool provides a structured layer of verification that complements runtime guardrails.
The separation of declaration checks from runtime execution is a deliberate design choice. A PASS result does not grant execution permission or enforce operating-system controls; those responsibilities remain with the operators. This clarity prevents confusion between successful preliminary checks and proof of operational safety, a distinction that many teams overlook.
The v0.5.0 release also introduces lifecycle-declaration consistency checks and task/action binding, showing Netsujo’s commitment to expanding the tool’s capabilities. For teams already using multi-agent frameworks, Agent Role Contracts offers a low-friction addition to their toolkit.
Verdict
Agent Role Contracts is an essential open-source tool for developers building reliable multi-agent AI systems with clear role boundaries. Its deterministic checks, offline operation, and concrete diagnostics make it a valuable addition to any workflow where agent permissions must be explicit and verifiable. The free MIT license and straightforward integration lower the barrier to adoption, while the thoughtful separation of checks from runtime enforcement sets a mature standard for the industry. For teams looking to add a layer of role verification to their AI-agent pipelines, Agent Role Contracts is a strong and timely choice.
Frequently asked questions
It is an MIT-licensed open-source tool from Netsujo that checks AI agent role declarations, task scopes, review assignments, and handoff rules for conflicts before execution.
No. It runs deterministic checks offline without any API key, AI model, or running agent.
It detects out-of-scope task requests, invalid review assignments (e.g., same role implementing and reviewing), and contradictory handoff declarations.
A PASS means the submitted declarations passed the built-in consistency checks. It does not grant execution permission or enforce operating-system access controls.
Clone the GitHub repository, run the three-minute demonstration, and integrate via the CLI or JavaScript API. It requires Node.js 22.5 or newer.
