Concepts
Organizations, MCP servers, tools, grants, runs and the audit log, and how they fit together.
Organization
Your team's space in sremcp. It holds the people and their roles, the MCP servers, the report schedules, the runs and the audit log. You can belong to several organizations and switch between them from the menu at the top of the sidebar. Each person has one role per organization. See Roles & permissions.
MCP server
A remote server that speaks the Model Context Protocol and exposes tools: for
example list_pull_requests on GitHub, query_prometheus on Grafana, or pods_log on Kubernetes. sremcp connects
to it over Streamable HTTP or SSE, from sremcp's servers, never from your browser.
A server has a status:
| Status | Meaning |
|---|---|
| draft | Just onboarded. The agent doesn't use it yet. |
| active | Live. The agent uses its approved tools, for the roles it's granted to. |
| disabled | Switched off by an SRE or admin. |
| error | The last check failed (unreachable, signed out, credentials refused). |
Tool
One capability of a server. sremcp lists a server's tools when you Test connection, and again on every health check. Each tool is pending until an SRE or admin approves it. If a server changes a tool's definition, that tool goes back to pending, so nothing new reaches the agent unreviewed. See Approve tools.
Grant
Which roles may use a server. Owners, Admins and SREs use every server. Managers, Developers, QA and Viewers use a server only if their role is granted it. A new server is granted to nobody. See Role access.
Run
One agent answer: a chat turn in Investigate, or a scheduled or manual report. A run records the question, the answer, the model, the number of tool calls, tokens and timing. See Runs.
Audit log
Every change (servers, approvals, grants, members, schedules) and every tool call the agent makes, with who did it and when. See Audit log.