Bring operational intelligence into Codex through MCP
Opsphere connects Codex to your infrastructure, observability, deployment, code and edge systems through MCP — combining direct operational queries with high-level, evidence-backed investigations.
From atomic operational tools to structured investigations and reusable context.
MORE THAN AN MCP TOOL CATALOG
More than an MCP tool catalog
Many MCP integrations stop at exposing individual actions. That is useful, but it makes every server look the same: a list of atomic calls.
Opsphere adds a structured operational layer behind those tools. Codex can run an atomic capability when a precise query is enough, or invoke a higher-level Opsphere workflow that combines several systems into one evidence-backed investigation.
Ops tools don't speak Codex MCP: your observability, deployment and paging stack was never built to return structured operational context to a Codex session over MCP. Opsphere sits between them and Codex.
Simple checks should stay simple
For a direct question — one Kubernetes resource, one endpoint, one deployment list — an atomic tool is the right tool. Opsphere keeps those available.
Multi-system troubleshooting needs structure
When a question spans deployments, errors, cluster state and edge checks, a flat list of tool calls forces Codex to reconstruct the whole investigation itself.
The intelligence layer belongs to Opsphere
Codex is the interface where engineers ask and act. The operational capabilities, scoping and guardrails live in Opsphere.
TWO WAYS IN
Atomic tools and high-level investigations
Opsphere does not force every question into an agentic workflow. Simple checks stay simple; complex investigations use higher-level context and structure.
Atomic tools
For direct operational questions: inspect a Kubernetes resource, check an HTTP endpoint, list deployments, query AWS inventory, search errors, inspect DNS and TLS state.
High-level investigations
For cross-system questions: investigate an outage, diagnose environment degradation, correlate deployment and error signals, and return structured evidence and verification.
ATOMIC TOOLS
Query individual systems directly
Codex can call individual Opsphere tools to query specific systems: Kubernetes resources, AWS infrastructure, Datadog errors, deployments, DNS / HTTP / TLS, repositories, CDN and edge state. A direct query when a precise answer is all you need.
EVIDENCE-BACKED INVESTIGATIONS
Run structured operational investigations
High-level Opsphere workflows can return a structured investigation to Codex instead of an opaque answer. Depending on the workflow, the result can include hypotheses, evidence, confidence, timeline, verification, observed facts, inferences and recommended next steps.
- Hypotheses
- Evidence
- Confidence
- Timeline
- Verification
Codex can reason over structured operational evidence instead of reconstructing the entire investigation from raw tool calls.
OPERATIONAL CONTEXT
Give high-level workflows reusable operational context
Opsphere keeps Operational Context separate from the model and the client. That context can include relevant relationships between services, environments, repositories and previous investigations.
High-level Opsphere workflows can use shared operational context, while atomic tools remain focused on individual systems.
SOURCE SYSTEMS AUTHORITATIVE
Operational context without copying your entire stack into the model
Opsphere does not load all of your telemetry or infrastructure into the Codex session. The source systems keep the operational data; Opsphere queries what is needed and returns structured findings and context.
Operational intelligence in Opsphere. Model interaction in Codex. Source data in the systems that own it.
READ-ONLY BY DEFAULT
Investigate without silently changing production
Codex can use Opsphere to inspect and investigate operational systems while Opsphere stays read-only by default. Recommendations can be structured, but they are not presented as autonomous changes executed by Opsphere.
EXAMPLE
Example: investigate a deployment regression from Codex
An engineer asks, in Codex: "Why did checkout-api start failing after the last production deployment?" Opsphere works through deployment evidence, observability evidence, Kubernetes state, network and edge checks, and relevant operational context.
- Probable cause
- Supporting evidence
- Alternative hypotheses
- Timeline
- Verification
- Recommended next checks
ONE LAYER, MANY INTERFACES
One operational layer across AI workflows
Codex is one of several interfaces from which Opsphere can be used. Cursor, Claude-compatible workflows, Codex and the Web Client can consume the same operational layer through the supported surfaces, keeping capabilities, context and guardrails consistent.
Different AI interface. Same Opsphere operational layer.
GUIDED SKILLS
Ready-made operational flows, not improvised prompts
Incident investigation
@incident-investigation
Multi-service outages and incidents → probable cause with evidence
Endpoint health
@endpoint-health
One hostname: DNS, HTTP, and certificate in one report
CI investigation
@ci-investigation
Failed pipelines and GitHub Actions (Professional+)
Post-mortem writer
@postmortem-writer
Structured RCA; optional save to Investigation Memory
Community: triage, endpoints, and post-mortems. Deep CI diagnostics on Professional, Team, and Enterprise plans.
KEEP EXPLORING
Related platform capabilities
GET STARTED
Connect Codex to your operational stack
Install the Opsphere Codex plugin, complete MCP OAuth once, and bring operational context and evidence-backed investigations into the Codex session.