> ## Documentation Index > Fetch the complete documentation index at: https://docs.meshpilot.app/llms.txt > Use this file to discover all available pages before exploring further. # Inbox > The single decision queue. One inbox, every agent, every brand. The Inbox is the cockpit's load-bearing surface. Every agent-drafted proposal — across every brand, every channel, every domain — lands here. There are no per-agent queues. There are no per-channel inboxes. One queue, filterable. ## The KPI strip ## How a proposal looks Each row tells you: * **From** — the agent that drafted it (Ads, Sales, Social, Creative, Voice, or SEO), as an attribution badge. * **Brand** — which brand it touches. * **What it proposes** — a one-line summary plus a payload preview. * **Why** — the agent's reasoning, citing the data it acted on. For audit-driven items, this is a structured rationale block: what was found, the evidence (e.g. "buy-box won 41% of the last 7 days"), and the recommended fix. * **Predicted impact** — where applicable (e.g. "saves \~₹12k/day at current spend"). * **Three actions** — Approve, Edit, Reject. Some items are **diagnoses, not mutations** — e.g. "do you agree with this anomaly read?" Approving those records your agreement and trains the model; it makes no campaign change. The dialog tells you plainly which kind of action you're approving, so "Approve" never means something you didn't expect. ## Filters Slice the queue by brand, source agent, domain (Ads / Sales / Social / Creative / Voice / SEO), and urgency. Filters persist across sessions per operator. ## Batch actions Select multiple rows and approve, reject, or expire them in one move. Useful when several low-risk proposals have stacked up (e.g. ten ad-set budget tweaks of the same shape). ## What happens after you decide * **Approve** → the orchestration engine writes the change back to the source-of-truth tool. The row moves to History. * **Edit** → tweak the payload (budget number, copy text, target list) before approving. * **Reject** → the row moves to History with your decision noted. Add an optional **note** explaining why — it's sent back to the agent so it can re-propose, and recorded with the decision. The agent reads the rejection signal into the Brain so it adjusts future proposals. * **Note on any decision** → both approve and reject accept an optional note, kept as a durable audit record of why you decided what you did. * **Expire** → proposals that have a window (e.g. time-sensitive bid changes) auto-expire if untouched. Expired rows appear in History as well. Inbox is not where you go to inspect work that's already running. For that, head to the relevant **Work surface** (Ads, Sales, etc.) or the **History** audit timeline. Go triage.