What is Infer and how does it secure AI Copilot interactions?

Infer is an MCP (Model Context Protocol) server designed to enable AI Copilot tools to securely interact with SSH and PostgreSQL resources. It secures these interactions by enforcing configurable policies, including command allowlists, query pattern restrictions, and resource limits.

What types of resources can AI Copilots access through Infer?

AI Copilots can securely interact with remote SSH servers for executing commands and PostgreSQL databases for querying. All access is governed by strict, policy-driven controls defined within Infer.

How does Infer manage access policies for SSH commands and PostgreSQL queries?

For SSH, Infer enforces command allowlists, output quotas, and execution time limits. For PostgreSQL, it applies statement pattern matching, row limits, and connection restrictions. Both also feature per-profile concurrency limits.

Can Infer facilitate AI model training orchestration?

Yes, Infer supports the orchestration of classifier training jobs. It leverages secure SSH profiles to enable controlled and policy-enforced execution of training scripts on remote infrastructure.

Infer

Name: Infer
Author: jackyxhb

0•

Enables AI Copilot tools to securely interact with SSH and PostgreSQL resources through policy-driven access controls.

Infer is a Model Context Protocol (MCP) server designed to bridge AI Copilot tools, such as GitHub Copilot and Cursor, with external computing and data resources. It facilitates secure and controlled access to SSH environments for command execution and PostgreSQL databases for querying. The server enforces strict policies, including command allowlists, output quotas, statement patterns, and concurrency limits, ensuring that AI tools operate within predefined boundaries. It also supports orchestration of tasks like classifier training via SSH profiles, making it a robust solution for managing AI-driven interactions with external systems.

主な機能

01Secure SSH command execution with configurable allowlists, output quotas, and runtime limits

02Policy-enforced PostgreSQL database querying with statement patterns, row limits, and connection restrictions

03Per-profile concurrency limits with cancellation support and progress notifications

04Orchestration of classifier training via defined SSH profiles

05Flexible configuration via JSON, supporting environment variables and secret file paths for credentials

060 GitHub stars

ユースケース

01Integrating AI Copilots with remote servers for code execution and system interaction

02Enabling AI tools to query and retrieve data from PostgreSQL databases securely

03Automating and managing machine learning model training jobs via AI assistant commands