Overview

The Overview is the landing page after login. It shows the current state of your workspace at a glance. At the top you see a time-based greeting (Good morning, Good afternoon, or Good evening) alongside the name of the selected agent.

​

Metrics

A KPI strip across the top shows six counters for the current workspace, in this order. Time-bounded counters use the latest 24h activity:

• Agents — Number of agents connected to your workspace.
• Sessions — Agent and chat sessions recorded in the latest 24h.
• API calls — Request count over the latest 24h.
• Graph nodes — Number of entities in the knowledge graph.
• Graph edges — Number of relationships in the knowledge graph.
• Brains — Total datasets (“brains”) in your workspace.

Counters show a placeholder dash while data loads and stay greyed out at zero.

​

Get started

The Get started section carries the subtitle Connect your AI agents to give them persistent memory and presents connection cards that walk you through linking an AI agent to your workspace. The cards are Claude Code, Codex, Openclaw, API / MCP, and Company Brain (upload). The Claude Code flow auto-detects a new session and marks the step done once it sees one. From the Company Brain card you can also upload files directly into a brain without leaving the dashboard.

​

Credits and vouchers

When credit usage is high, a banner appears on the Overview:

• At ≥90% credits used, a low-credit warning is shown.
• At 100%, the banner notes that agent requests may fail.

Owners see a Top up credits → action that links to Billing. Non-owners see Ask the workspace owner to top up. A single banner shows below this warning. Normally it is the voucher banner (You have a voucher? Redeem it here., linking to Billing). When your remaining token balance falls below 1 USD, it is replaced by a red low-balance banner reading Your Token Balance is below 1 USD. You can recharge credits here. The low-balance banner also links to Billing and stays visible even after the credit-usage warning above has been dismissed. These credit-usage, voucher, and low-balance banners appear only in cloud mode. They are hidden in the local, open-source UI, where credits and billing do not apply.

​

LLM API key (local mode)

When you run Cognee locally (the open-source UI, launched in local mode), the Overview checks the backend settings and, if no LLM API key is configured, shows a warning banner reading No LLM API key configured — Cognee can’t process uploads until you add one. with an Add your API key now → action. Clicking it opens an Add your LLM API key modal where you can paste your key. Saving applies the key to the running backend right away and reloads the page, so uploads and cognify work immediately — no .env edit or restart required. The banner only appears in local mode; it is not shown in the cloud.

​

Cognee Terminal

The Search your memory terminal lets you query your knowledge graph directly from the dashboard. If a brain is selected, the query is scoped to that brain; otherwise, it searches across the visible brains. Results are returned inline using the same recall mechanism as the dedicated Search page.

​

Quick actions

• Connect with Python SDK — Opens a modal with step-by-step setup instructions: install the package (pip install cognee), get your API Base URL and API key, and connect your agent.
• Upload files — Uploads files directly into the currently selected dataset without leaving the dashboard. Accepted file types are .pdf, .csv, .txt, .md, .json, and .docx. If multiple brains exist and none is selected, an Upload to which brain? picker appears first. When processing finishes, a Knowledge graph built modal offers next steps: Search your data (opens Search), Inspect the knowledge graph (opens the brain in Datasets), and Explore the knowledge graph (opens the Mindmap).
• What you can build — A scrollable row of use-case cards illustrating what Cognee can do with your data: A second brain, Sales & deal intelligence, Investment & research, Docs & manuals, and Memory for coding agents.

​

Memory Activity

The Memory Activity log is a live, evidence-oriented feed of every search run against your memory — both by your connected agents and by you — sorted by recency. Each actor is shown with a stable colored dot so distinct agents are easy to tell apart. Filter tabs narrow the feed:

• all — every event.
• mine — only the queries you typed yourself.
• agents — activity from other actors (connected agents and the system).
• searches — search/recall events only.
• errors — failed queries only.

Your own typed queries always remain visible on every tab except agents, so a query you ran never silently disappears from the log. Every row carries an outcome badge so a recall that returns nothing is explained rather than left silent:

• hit (✓, green) — the recall returned results.
• empty (∅, yellow) — the recall ran but found nothing.
• error (✗, red) — the query failed.
• running (spinner) — the query is still in progress.

Click any row to expand it and see what was searched and why it answered. The most recent settled query expands automatically; once you collapse or pick a card manually, it stays as you left it. For full pipeline run history and trace details, see the Sessions page.

​

Give feedback

A Give feedback button is pinned to the bottom of the sidebar (above Billing) on every page of the app. Clicking it opens a short form; submitting sends your message straight to the Cognee team, tagged with your current workspace and the page you were on for context. You must be signed in to send feedback.