MCP server
Connect Claude, ChatGPT, Codex and other AI agents to uptime and security context, with local Git remediation skills and approval before deployment.
Statance exposes a remote Model Context Protocol server so AI agents can work with your monitoring data.
- URL:
https://api.statance.com/mcp - Transport: Streamable HTTP
- Plugin auth: OAuth with read-only
statance:readaccess; sign in to Statance to approve the connection. - Existing API clients:
Authorization: Bearer YOUR_API_KEY. Account keys retain monitor-management access; OAuth plugins do not.
Claude and ChatGPT plugins
Use the same remote MCP URL in Claude or ChatGPT and choose OAuth. Sign in to Statance, review the client and permissions, then allow read access. Revoke individual clients from AI connections.
ChatGPT developer mode supports adding a custom remote MCP server with OAuth. Statance supports dynamic client registration (DCR), PKCE, refresh tokens and revocation. Public directory listing is a separate review process; a custom connection does not require Statance to be listed.
Download the Statance plugin for private testing. It includes Claude manifests, a portable OpenAI manifest and three skills: triage, local Git patching and verification. Extract it, then test in Claude Code from your application repository:
claude --plugin-dir /absolute/path/to/statanceUse /mcp to connect and approve access. For a supported private OpenAI plugin import, use the portable package. For a public OpenAI listing, register the MCP server in the Developer Portal, attach the skills and complete its review process. Importing a Claude archive alone does not register the remote server.
Prepare patches with your agent
Ask: “Review my Statance security findings and uptime. Prepare a tested patch in my local Git repository for finding ID … and ask before deployment.” Security detail follows your plan: Pro or Business is required.
The agent uses observed versions, advisory references, remote lockfile paths and scan timestamps to investigate. It must verify which local repository deploys the affected application, inspect its instructions and latest main, prepare an isolated patch and run relevant checks. Ordinary chat without local Git access can investigate and prepare a handoff.
Before deployment, the agent must show the exact revision and target, tests, migration or restart impact, rollback and verification plan, and ask for explicit approval. Auto-deploying pushes or merges, OS upgrades, restarts, reboots and firewall changes also require approval. Skills guide behavior; client permissions and deployment approval gates must enforce this boundary. Statance MCP provides no shell or deployment tool.
Claude Code with an account API key
claude mcp add --transport http statance https://api.statance.com/mcp --header "Authorization: Bearer YOUR_API_KEY"Other API-key clients
{
"mcpServers": {
"statance": {
"url": "https://api.statance.com/mcp",
"headers": { "Authorization": "Bearer YOUR_API_KEY" }
}
}
}Tools
| Tool | What it does |
|---|---|
list_servers | Servers with status, OS and latest metrics |
get_server_metrics | Metric history for a server over a time range |
get_top_processes | Latest top processes on a server |
list_monitors | Monitors with status and last response time |
get_monitor_checks | Recent check results for a monitor |
create_monitor | Create an uptime monitor |
pause_monitor / resume_monitor | Pause or resume a monitor |
list_incidents | Open or recent incidents with AI summaries |
list_alert_rules | Server alert rules |
investigate | Look up a domain or IP: registration, DNS, email security, IP owner, subdomains and a live check |
get_security_findings | Prioritised security findings with stable IDs, affected versions and pagination; follow next_offset until null (details on Pro and Business) |
get_security_finding | Finding evidence, advisory links, fixed versions, scan health and affected remote lockfile paths |
get_server_security_context | OS/kernel, software, posture, updates, containers, listeners and observed lockfile inventory for an exact server ID |
Example prompts
- “Which of my servers is closest to running out of disk?”
- “Add an HTTP monitor for https://staging.example.com every 5 minutes.”
- “Summarise the incidents from the last 24 hours.”
OAuth tools are read-only and scoped to the connected account. Account API keys can also manage monitors. Both follow plan limits. Use the remediate_security_finding MCP prompt where supported, and check scan timestamps before treating a missing finding as fixed.
Next: AI assistant →