Causely, the causal intelligence layer for AI agents operating production systems, today announced upgrades to its MCP server based on how background agents use it in production. The release makes agents faster to deploy, easier to trust, and cheaper to run, and reflects months of learning from how agents use it at customers including Unity and Quantum Metric and in Causely’s own environment. Causely’s MCP server is now listed in the GitHub MCP Registry, and new working examples of background agents are available in Causely’s open source repositories. The announcement coincides with AGNTCon + MCPCon North America 2026, where Causely is a sponsor.

Background agents are taking on a growing share of the work that keeps applications performant. They run on a schedule or a trigger, investigate on their own, and hand back a diagnosis or a fix. Teams building them keep hitting the same three problems: the agent takes weeks of prompt engineering before it uses tools well, when it acts nobody can easily check what it based the decision on, and it burns tokens rediscovering the state of the system on every run.

This release addresses each of those.

Faster to deploy. Causely’s MCP server now carries the guidance an agent needs to use its tools well, so a new agent is productive on first connection instead of after weeks of tuning.

Easier to trust. Every agent action can be traced back to the diagnosis and evidence it was based on, and teams can see how their agents are using Causely over time. A pull request or Slack message from an agent now comes with a trail a human can check.

Cheaper to run. Agents can start from what is wrong and what recently changed, delivered as structured results, rather than scanning the environment to work it out. Causely’s published benchmark across 72 experiments showed agents with this context reached 100% fault detection accuracy while using 48% fewer tokens and returning answers 63% faster.

Unity runs background agents on Causely’s MCP server across thousands of services. Multiple teams have integrated Causely into their agents, including an internally developed AI SRE and agents built on Claude, HolmesGPT, and Hermes. Causely is the first tool those agents call when investigating an incident, and the same agents run continuous scans of the environment to identify reliability blind spots. Quantum Metric connects Causely to the agent its engineers work with while investigating problems, so the first answer an engineer gets back already includes the cause and what it affects.

“Our engineers work across thousands of services. With Causely behind the agent they use, they get a better answer, faster, and one they can act on with confidence,” said Glenn Trattner, COO of Quantum Metric.

“We have spent the past year seeing how background agents run against Causely, at customers and in our own environment, and the lesson was consistent. An agent is only as good as its understanding of the system it operates on. This release puts more of that understanding into the MCP server itself, so agents are productive sooner, teams can trust what they act on, and every run costs less,” said Yotam Yemini, CEO of Causely.

The updates are available now to all Causely customers at api.causely.app/mcp and work with any MCP-compatible client, including Claude Code, Cursor, Codex, Hermes, and HolmesGPT. Background agent examples and Causely MCP Skills are open source at github.com/causely-oss. Causely will be at booth 80 at AGNTCon + MCPCon North America, October 22 to 23.

About Causely

Causely provides the causal intelligence layer for AI agents operating production systems. Its continuously maintained causal model turns telemetry into structured, machine-consumable understanding of cause and effect, delivered through MCP, so agents diagnose accurately, use fewer tokens, and act sooner. Causely is a Silver member of the Linux Foundation, CNCF, and the Agentic AI Foundation. Learn more at causely.ai.

Media gallery

About The Author
Business Wire