inkog-mcp

Inkog MCP Server

Inkog supports developers building and reviewing AI agents through MCP clients such as Claude Code and Cursor. It scans agent code, checks AGENTS.md governance, explains findings, and audits MCP integrations and multi-agent delegation.

Server facts

inkog-mcp

Project

Author
inkog-io
Website
inkog.io
License
Apache-2.0

Tech

Runtime / language
node
Transports
Not specified
Remote available
No
Authentication
Not specified

Activity

GitHub stars
3
Last commit
Data as of

Overview

AI agent development with Inkog centers on a scan, explain, fix, and rescan workflow inside an MCP client. Developers can inspect files or directories for problems such as prompt injection, runaway loops, unrestricted tool use, and data leakage, then request explanations and remediation examples for individual findings. Governance checks compare AGENTS.md statements with implementation behavior, including whether declared approval steps or access limits appear in the code. A separate deep analysis tool targets complex vulnerabilities and logic errors that pattern-based checks may miss.

Integration reviews extend beyond the agent’s own code. Inkog includes tools for examining MCP servers supplied through a registry or repository URL, plus a scanner for SKILL.md packages and agent tool definitions. Multi-agent audits inspect delegation for loops, privilege escalation, unintended information sharing, and unauthorized transfers of work. For documentation and audit preparation, compliance reports relate code findings to frameworks including the EU AI Act and NIST AI RMF. Its Machine Learning Bill of Materials tool inventories AI components and dependencies, with CycloneDX and SPDX output options.

Related MCP servers