InGravity is an open-source project designed to help developers control, secure, and govern AI agent actions.
As AI agents gain the ability to interact with APIs, execute workflows, modify files, and perform sensitive operations, developers need a reliable way to define which actions are allowed, which must be blocked, and which require human approval.
InGravity introduces a policy-based control layer that evaluates agent actions before execution and returns one of three decisions:
- Approved — The action is permitted by the configured rules.
- Blocked — The action violates a policy or must not proceed.
- Pending approval — The action requires human authorization.
Current capabilities
- Configurable rules and risk-based action evaluation.
- API for checking AI agent actions.
- Action history and decision tracking.
- Human approval workflows.
- Dashboard for managing rules and reviewing actions.
- n8n integration through the InGravity Guard node.
Our vision
We want to make AI agent governance accessible to independent developers and small teams, without requiring complex enterprise infrastructure.
Our long-term direction includes stronger runtime enforcement, identity and permission controls, emergency shutdown capabilities, MCP connectivity, and deeper security checks for AI-driven workflows.
InGravity is being developed as an open-source project, with the goal of making AI agents safer and more controllable while preserving developers’ flexibility.
We’re looking for developers interested in AI agent security, policy enforcement, MCP, automation, and open-source security tools to explore the project, test integrations, share feedback, and contribute.
GitHub: https://github.com/Tchesseludovic17-cloud/ingravity
InGravity is an early-stage project. Current capabilities should be distinguished from planned features, and integrations must enforce InGravity’s decisions for the control layer to be effective.
Top comments (0)