Introducing Antimetal for Coding Agents

Matt Casey

April 16, 2026

Your coding agent can write code, generate tests, plan architecture, and deploy. But the moment something breaks in production, it can't help. You're back to context-switching between Datadog, CloudWatch, and PagerDuty, piecing things together yourself. The agent that helped you ship cannot help you put the fire out.

Today, we're launching the Antimetal MCP. A single connection that gives your coding agent runtime context across your entire infrastructure and observability stack, enabling you to go from issue to fix without leaving your editor or CLI.

Introducing the Antimetal MCP

Antimetal pulls from 50+ integrations — Datadog, CloudWatch, Grafana, PagerDuty, and more — so you don't have to manage a separate MCP for each tool.

But we don't just aggregate your data. We've built our own representation on top of those sources that maps the topology, evolution, causal behaviors, and semantic labels of your system.

This matters because individual MCPs weren't designed for this multi-faceted workflow, where you're trying to understand and diagnose a large system holistically. Each tool returns a different payload format, each requires its own tool call, and the agent has to reconstruct the relationships between these signals every time. And the more tools you load into context, the worse the output.

Antimetal normalizes all of that into a small set of tools that give your agent a structured understanding of your system, not just raw signals. The result is better retrieval, faster investigations, and more accurate fixes.

Automated workflows through skills

With a simple slash command, you can leverage two powerful workflows using our MCP:

/investigate: entry point for any production problem. Describe what's going wrong, and Antimetal investigates in the background.
/fix: get a fix applied directly to your codebase.

6 tools

Tool	What it does
`investigate_issue`	Trigger an async investigation. Describe symptoms, Antimetal investigates using your runtime data
`get_issue_report`	Pull results once ready: root cause analysis, causal graph, timeline
`get_issue_fixes`	Get remediation actions: code changes, CLI commands, instructions
`search_issues`	Find issues across your environments
`get_artifact`	Fetch supporting evidence: metrics, traces, logs, events, files, service topology
`ask`	Natural language query. Delegates to Antimetal's AI agent, not a raw model call. Backed by tools and your observability data

An example: debugging a latency spike

Datadog alerts fire: p99 latency on /api/checkout has tripled in the last ten minutes. Antimetal picks up the alert and automatically starts an investigation.
You type: /investigate the latency spikes on the checkout service. Antimetal already has results.
The investigation pulled metrics from Datadog and traces from CloudWatch, built a causal graph, and traced the spike back to a recent deploy. A PR merged earlier that morning updated the ECS task definition and dropped the DB_POOL_SIZE environment variable. Every request is now opening a new database connection.
You type /fix. Your agent takes Antimetal's proposed remediation and applies it to your codebase: adding the missing variable back to task-definition.json and updating the Terraform module that generates it.
You review the changes, commit, push, deploy. Latency drops back to baseline.

Get started

The server is live at mcp.antimetal.com and works in Cursor, Claude Code, VS Code with GitHub Copilot, and any editor that supports the Model Context Protocol.

If you're an Antimetal customer, install the plugin from github.com/antimetal/skills and bring it to your coding agent today. If you're not, book a demo and we'll get you set up.

You can find additional info at docs.antimetal.com/connect.

Matt Casey

April 16, 2026