Agentguard47
Expose coding-agent traces, spend alerts, and budget health to your IDE agent through a read-only AgentGuard MCP feed.
Overview
AgentGuard47 is an MCP server for the Operate phase that read-only queries coding-agent traces, alerts, costs, usage, and budget health from AgentGuard.
What is this MCP server?
- Read-only MCP server—query traces and alerts without mutating AgentGuard state
- Surfaces coding-agent costs, usage, savings, and budget health metrics
- npm package @agentguard47/mcp-server v0.2.2 with stdio transport
- Requires AGENTGUARD_API_KEY; optional AGENTGUARD_URL for non-default API base
- Purpose-built for agent observability, not generic log tailing
- Package @agentguard47/mcp-server version 0.2.2
- Transport: stdio via npm registry
- Required secret env: AGENTGUARD_API_KEY
Community signal: 3 GitHub stars.
What problem does it solve?
Solo builders lose visibility into how much their coding agents spend and misbehave until bills spike or sessions fail without a clear trace.
Who is it for?
Builders already on AgentGuard who want MCP-native cost and trace checks inside Claude Code, Cursor, or similar stdio MCP hosts.
Skip if: Teams without an AgentGuard account, or anyone who needs write/control plane actions from the agent.
What do I get? / Deliverables
You can ask your agent for AgentGuard-backed usage, cost, and alert summaries while coding, using a scoped read API key.
- Trace and alert summaries pulled into agent conversations
- Cost, usage, and savings views aligned with AgentGuard records
- Budget health snapshots for go/no-go decisions on heavy agent runs
Recommended MCP Servers
Journey fit
Cost and trace visibility matter after agents run in production-like daily workflows, which maps to Operate rather than first-time build. Monitoring is the canonical shelf for usage, alerts, and budget signals you watch while keeping agent-assisted shipping sustainable.
How it compares
Read-only agent observability MCP, not an application error tracker or infrastructure metrics stack.
Common Questions / FAQ
Who is AgentGuard47 MCP for?
Solo and small teams using AgentGuard to watch coding-agent spend and reliability who want those metrics available inside their MCP-enabled IDE agent.
When should I use AgentGuard47 MCP?
Use it during daily Operate workflows when reviewing budgets, investigating agent alerts, or correlating traces with a shipping week.
How do I add AgentGuard47 MCP to my agent?
Install @agentguard47/mcp-server from npm, set AGENTGUARD_API_KEY (and optional AGENTGUARD_URL), and register the stdio MCP server in your client configuration.