io.github.CSOAI-ORG/agent-negotiation-mcp

The materials explicitly state that no keys or environment variables are required, and there is no indication of API keys, OAuth tokens, or sensitive local credentials being requested; based on the available facts, credential exposure appears low.

It is known to access the remote endpoint api.meok.ai, so prompts, negotiation context, or task data may be sent to that service. The endpoint appears related to the stated MEOK AI Labs function, with no obvious unrelated exfiltration red flag, but the missing README leaves the scope and content of transmission unclear.

System checks indicate that the tool executes code/spawns processes, which is common for MCP tools but still creates a local execution surface. The materials do not clarify whether it can further invoke a shell, pass arbitrary subprocess arguments, or call external programs, so the permission boundary is not fully clear.

The provided materials do not specify what local files, session data, or other resources it can read or write, so compliance with least-privilege cannot be confirmed. There is no explicit evidence of overbroad access, but the lack of documentation prevents full auditing of its data access scope.

There are positive supply-chain signals: official registry listing, an open-source repository, and updates within the last year, all of which reduce risk. At the same time, the missing README, undeclared license, and very low community adoption (0 stars) weaken auditability and external validation, making caution more appropriate than high risk.