STATUSCARD is the system’s agent observability extension. Its job is not just to show whether an agent exists, but to turn the current OpenClaw topology into live Discord cards that reflect how the agent layer is actually behaving. It reads the configured agents from openclaw.json, groups them by shared workspace, chooses one root card per workspace, and then discovers subagents dynamically instead of relying on a hardcoded list. That means the extension naturally adapts when new subagents are added or removed from the configuration. It also inspects session ownership using Discord channel and thread bindings, session-key prefixes, and shared workspace session files, which lets it map activity back to the correct root agent or subagent even when session shapes differ. In practice, this makes statuscard a compact live map of the agent system: model, think level, context usage, main cost, aggregated subagent cost, active session count, and per-subagent activity all appear together in one operational view. It exposes /statuscard help, /statuscard agents, /statuscard init, /statuscard subagents <agentId>, and removal commands, so it functions as both a dashboard and a control surface for card lifecycle management. A useful technical detail is that it prefers configured model IDs and configured thinking levels when available, but it can fall back to runtime .jsonl events such as thinking_level_change, which makes the cards more accurate than a purely static config snapshot. Its main references are card-context.md, index.js, and statuscard-core.js.

CODEXSTATUS is narrower in scope but equally important: it is the quota and renewal monitor for OpenAI/Codex usage. Where statuscard observes agents, codexstatus observes consumption windows. The current implementation is fetch-only, meaning it does not invent limits or compute arbitrary budgets; instead, it tries to reflect real usage data from sources it can verify. Its preferred source is the same live usage snapshot used by openclaw status –json –usage, and it can also query the OpenAI usage endpoint at /v1/organization/usage/completions. If those sources do not yield usable daily or weekly % left values, it falls back to parsing recent Discord messages in the same channel and then recent agent session logs, looking for quota phrases such as “Daily usage: 79% left” or natural-language equivalents. That fallback design is one of the most distinctive characteristics of the extension: it is resilient, source-layered, and explicitly protects itself from re-ingesting its own card output. On the Discord side, it renders a compact OpenAI Token Status card with separate daily and weekly sections, progress bars, used percentages, and renewal timing, with daily renewal expressed in hours and weekly renewal in days. It also keeps channel-specific memory in state.json, supports /codexstatus, /codexstatus refresh, /codexstatus show, /codexstatus help, and /codexstatus remove, and is designed to be refreshed by its companion updater script. A key limitation, documented in the extension itself, is that public usage endpoints do not always expose Plus-style % remaining directly, so the card may legitimately show N/A until discoverable values appear in a trusted source. Its main references are codexstatus-context.md, index.js, and codexstatus-core.js.

Together, these two extensions form the custom monitoring layer of the system. statuscard answers “what are the agents doing right now, and how is the workspace tree behaving?”, while codexstatus answers “what is the current OpenAI/Codex usage situation, and when do the limits renew?” The first is topology-and-session observability; the second is quota-and-renewal observability. That division is clean, practical, and very consistent with the broader role of the status-agent in this installation.