AEGIS Governance — six-gate decision evaluation with hash-chained audit trails (stdio via PyPI + hosted Streamable HTTP)
@undercurrentai
关于 AEGIS Governance — six-gate decision evaluation with hash-chained audit trails (stdio via PyPI + hosted Streamable HTTP)
AEGIS Governance MCP server — six-gate quantitative decision evaluation with hash-chained audit trails for AI agents. pip install aegis-governance
基本信息
配置
工具
10Full six-gate evaluation of a proposal; returns PROCEED/PAUSE/HALT/ESCALATE with per-gate scores and rationale
Fast risk screen for a proposed change
Current gate threshold configuration
Domain-specific guidance for deriving gate parameters (e.g. `cicd`)
Record a proposal for later verification
List recorded proposals
Verify recorded proposals against outcomes
List past governance decisions
Fetch a specific decision with full audit detail
Hash-chain audit integrity status
概览
What is AEGIS Governance?
A quantitative governance server for AI agents and engineering decisions. It evaluates proposals through six gates—Risk, Profit, Novelty, Complexity, Quality, Utility—and returns a structured decision (PROCEED, PAUSE, HALT, or ESCALATE) with confidence scores, rationale, and a hash-chained audit trail. Designed for teams that need verifiable, auditable decision records in line with standards like NIST AI RMF and EU AI Act Annex IV.
How to use AEGIS Governance?
Install via pip with the MCP extra: pip install "aegis-governance[mcp]". Then add the server to any stdio MCP client (Claude Code, Cursor, Windsurf, VS Code) using the command aegis-mcp-server. For the hosted server with all ten tools, use a Streamable HTTP transport with a free API key from portal.undercurrentholdings.com and the URL https://mcp.aegis.undercurrentholdings.com/mcp. Alternatively, use the Python SDK directly for local evaluations.
Key features of AEGIS Governance
- Six quantitative gates with calibrated thresholds and drift detection
- Hash-chained audit trail for tamper-evident decision records
- Local sandbox mode with 10 evaluations per day, no signup
- Hosted server with ten tools and free Community tier (100 evaluations/month)
- Five integration surfaces: MCP, Python SDK, REST API, CLI, GitHub Action
- NIST AI RMF and EU AI Act Annex IV artifact generation
Use cases of AEGIS Governance
- Evaluate an AI agent's proposed action before it executes
- Perform a fast risk screen on a change to infrastructure or code
- Record and later verify engineering proposals with an immutable audit trail
- Generate compliance-ready artifacts for regulated environments
FAQ from AEGIS Governance
Do I need an API key to use AEGIS Governance?
No, the local stdio server runs in sandbox mode with no account needed (10 evaluations per day). A free API key unlocks the hosted server with the full tool surface and higher limits.
What are the runtime requirements for AEGIS Governance?
Python >= 3.10 is required. The local server is installed via pip; no additional services or accounts are necessary in sandbox mode.
What are the evaluation limits?
The local sandbox limits evaluations to 10 per day. The hosted free Community tier provides 100 evaluations per month with no credit card required. Paid tiers for higher capacity are available.
How do I get an API key for the hosted server?
Sign up at portal.undercurrentholdings.com using GitHub or Google sign-in. A free API key is provisioned automatically and can be used immediately with the Streamable HTTP endpoint.
Does AEGIS Governance support both stdio and Streamable HTTP transports?
Yes. The local server uses stdio and is compatible with Claude
其他 分类下的更多 MCP 服务器
Core Philosophy: Connect, Unify, Respond
mindsdbDelegate anything. It comes back done.
MCP Toolbox for Databases
googleapisMCP Toolbox for Databases is an open source MCP server for databases.
MCP Registry
modelcontextprotocolA community driven registry service for Model Context Protocol (MCP) servers.
Mcp
browsermcpBrowser MCP is a Model Context Provider (MCP) server that allows AI applications to control your browser
Awesome Mcp Servers
punkpeyeA collection of MCP servers.
评论