Opsphere meets Codex at the agent layer
Bring live infrastructure context into OpenAI Codex — topology signals, incident evidence, and guided remediation skills surface in Codex CLI or ChatGPT desktop before you leave the session.
Guided skills for incidents, endpoint health, CI, and post-mortems — without jumping between dashboards.
Connection Hub: link multiple client Opsphere workspaces to one Codex login. Open a server-side work context, run scoped tools with context_id, and switch clients without re-authenticating.
THE AGENT-NATIVE PROBLEM
Ops context should travel with the agent — not live in another tab
Engineers work in Codex, but production truth still lives in dashboards, terminals, and runbooks. Every switch costs time when an incident is already burning.
When something breaks, investigation starts in the Codex agent session — yet confirming blast radius still means abandoning the agent to check logs, CI pipelines, and stale playbooks.
Ops Tools Don't Speak Codex MCP: Your observability stack was never designed to stream structured ops context directly into Codex over MCP.
No Operational Context in Codex
Codex has no built-in view of what is failing in production. Engineers leave the agent mid-investigation to chase dashboards, stretching every MTTR.
Incident-to-Remediation Handoff Is Broken
Pager alerts, Slack threads, and runbook tabs fragment the workflow. Opsphere returns root-cause evidence and next steps where the agent already works.
Ops Tools Don't Speak Codex MCP
Most observability products stop at dashboards. Opsphere exposes a production-grade MCP gateway built for infrastructure teams — with a Codex-native plugin and OAuth path.
HOW OPSPHERE INTEGRATES
Live in Codex in a few minutes
Install the Opsphere Codex plugin, complete MCP OAuth once, and start with guided skills — no agents to deploy and no infrastructure changes required.
Install Codex Plugin
Add the Opsphere plugin via `codex-install.sh` or the ChatGPT desktop plugin directory. Manifest and MCP metadata ship with `.codex-plugin/plugin.json` — zero local sidecars.
Connect & OAuth
Authenticate to the Opsphere MCP gateway via Codex OAuth. Link multiple client workspaces with the Connection Hub and switch environments effortlessly using `context_id`.
Invoke Guided Skills
Run `@incident-investigation`, `@endpoint-health`, or `@ci-investigation` inside your Codex session. Over 310+ operational tools execute securely without switching context.
TECHNICAL BENEFITS
Built for how Codex and MCP actually work
Opsphere's Codex integration is engineered around the Model Context Protocol — with a Codex-specific plugin manifest, OAuth client, and script setup, not a Cursor config pasted into another tool.
Native MCP Integration
Opsphere speaks MCP natively to Codex. Direct, spec-compliant traffic to the Opsphere gateway with automated setup via `codex-mcp-config.sh` and CI validation.
Codex CLI and ChatGPT Desktop
Validated for Codex CLI end-to-end and ChatGPT desktop plugin install. Same gateway, same skills, client-specific OAuth identity for Codex.
Read-Only, Secure By Design
Opsphere observes your stack — it does not rewrite production by default. Transit is encrypted; credentials never land in repo files.
Fast Context for Agents
Topology, signals, and incident history load quickly so Codex can reason without stalling the session.
Skill-Aware Operations
Guided `@skills` expose structured workflows — outage triage, endpoint health, CI investigation, and postmortems — so prompts stay purposeful instead of improvised.
Codex Integration Specifications
- Context load time
- <800ms
- MCP integration
- Native
- Supported clients
- Codex CLI · ChatGPT desktop
- Install footprint
- Codex plugin
- Default permissions
- Read-only
- Operational tools exposed
- 310+
- Primary clients
- Codex
- Transport
- Remote MCP · OAuth
- Agent context format
- Structured evidence
- Guided skills
- 4+ · Read-only · @skills
GUIDED SKILLS
Ready-made operational flows, not improvised prompts
Incident investigation
@incident-investigation
Multi-service outages and incidents → verdict + 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 memory
Community: triage, endpoints, and post-mortems. Deep CI diagnostics on Professional, Team, and Enterprise plans.
WORKFLOW EXAMPLE
From alert to fix: stay in the Codex session
Agent-Native Incident Flow
Here's what Opsphere does when a production anomaly hits while you are working in Codex.
Anomaly Detected On Prod API Cluster
Request error rate spikes to 8.4%; correlated with a new deployment 12 minutes prior.
Opsphere Surfaces Context In Codex
Root cause, affected services, and blast radius return through MCP — no dashboard detour required.
Engineer Triggers Remediation From The Session
Remediation guidance is generated in-chat; on-call coordination continues without leaving Codex.
Incident Resolved — Minutes, Not Context Switches
Cluster healthy; post-incident summary captured; the engineer never abandoned the agent session.
GET STARTED TODAY
Bring operational intelligence into Codex.
Join engineering teams who keep ops context inside the agent session — and cut incident response time without leaving Codex.