User Guide

Interface Overview

CogmemUI's interface is divided into four main areas:

• Header — Model selector, search, memory access, settings, theme toggle, and new-chat button
• Sidebar (left) — Conversation list with search, import/export, and infinite scroll
• Chat Area (center) — Message thread, input bar, attachment previews, pinned tools, and minimap
• Panels (right) — Collapsible panels for Files, Memory, Canvas, Artifacts, Think, and Git

The layout is fully responsive. On mobile, the sidebar and panels become slide-in drawers accessible via buttons in the header.

Managing Conversations

Creating a new chat

Click the + button in the header, press Ctrl + N, or type /new. The welcome screen appears with starter templates you can click to jump right in.

Searching conversations

Use the search bar at the top of the sidebar to filter by conversation title. Results update as you type.

Multi-tab conversations

CogmemUI supports tabbed conversations. Open multiple conversations side-by-side with the tab bar above the chat area. Click a tab to switch, or right-click for options.

Exporting conversations

Export conversations in multiple formats:

• /export — Export the current conversation as a JSON file
• /pdf — Export the current conversation as a formatted PDF document

Use the Import button in the sidebar to load previously exported JSON conversations.

Sharing conversations

Type /share to generate a shareable link for the current conversation. Share it with teammates or anyone you'd like to see the conversation.

Working with Ai Models

Switching models

Click the model selector in the header or type /model to switch models. Models are grouped by provider. Only providers with a configured API key are shown. You can switch models mid-conversation — each message can use a different model.

Auto Router

Select Auto in the model selector. CogmemUI will analyze each message and automatically route it to the optimal model based on task type (code, creative, quick-answer, reasoning, translation, math). The router also respects your budget settings, preferring cheaper models when you're near your spending limit.

Compare all models

Type /compare followed by your prompt to send the same message to all your configured models simultaneously. See how different models respond to the same question side by side.

Extended thinking

Some models support extended thinking (chain-of-thought reasoning). Toggle it with /think or the brain icon in the header. When active, the Think Panel on the right side shows the model's reasoning process in real-time with a token counter.

Input & Attachments

Sending messages

Type in the input area and press Enter to send. Use Shift + Enter for a new line without sending.

File attachments

Click the paperclip icon or drag and drop files into the chat. Supported types:

• Images (JPEG, PNG, GIF, WebP) — automatically resized to save API costs
• Documents (PDF, DOCX, XLSX)
• Text files (TXT, CSV, Markdown, HTML, JSON)

Voice input

Click the microphone icon next to the send button to dictate your message. Voice input uses the browser's Web Speech API and shows a real-time transcript. Available in Chrome and Edge.

Voice conversation mode

Type /voicechat to enter full voice conversation mode. CogmemUI listens continuously, auto-sends after silence, reads the Ai response aloud, then resumes listening. Tap to interrupt at any time. Toggle voice output only (without listening) with /voice.

Screenshot capture

Click the screen icon to capture a screenshot of any window or tab and share it with the Ai. Available in Chrome, Edge, and Firefox.

Persistent Memory

CogmemUI integrates CogmemAi for persistent Ai memory. When enabled, the Ai automatically saves and recalls important facts: your preferences, project architecture, bug fixes, decisions, and more.

Viewing memories

Click the brain icon in the header or type /memory to open the Memory Panel. Browse your memories, search by keyword, or filter by type and category.

How memories work

• Memories persist across all conversations — not just the current one
• The Ai automatically saves important context as you chat
• Relevant memories are recalled and injected into conversations where they’re useful
• You can manually save, edit, or delete memories from the Memory Panel

The Intelligence Engine

CogmemAi’s Intelligence Engine makes your memory smarter over time — automatically, with no effort from you:

• Self-improving recall — memories that help you rank higher automatically; unused ones fade naturally
• Auto-linking knowledge graph — related memories connect automatically when saved, building a web of relationships
• Contradiction detection — when recalled memories conflict, CogmemAi flags the contradiction so you catch stale information
• Query synthesis — ask a question and get one coherent answer synthesized from all relevant memories, not a list of matches
• Context-aware ranking — tell CogmemAi what you’re doing (debugging, planning, reviewing) and the right types of memories surface first
• Cross-project intelligence — patterns that appear across multiple projects automatically promote to global scope
• Auto-Skills — your corrections become behavioral rules that teach your Ai how to work with you. Correct once, and it never happens again

Memory tools

The Ai has access to five memory tools it can use automatically during conversations:

• Save memory — Store facts, preferences, decisions, and important details
• Recall memories — Search stored memories for relevant context
• Get project context — Load top memories for the current project
• List memories — Browse stored memories with pagination
• Delete memory — Remove a specific memory when asked

Getting a CogmemAi key

Sign up for a free key at hifriendbot.com/memory, then add it in Settings > API Keys.

Server Tools

Every Ai model in CogmemUI can use powerful server-side tools. The Ai decides when and how to use them based on your request — just ask naturally.

Web & search tools

Image generation

File operations

Content & system

Canvas & Artifacts

Canvas editor

The Canvas is a side-by-side code/content editor with live preview. Open it with /canvas or from the panel buttons on the right. Use it to:

• Edit code with syntax highlighting
• Preview HTML, SVG, and Markdown in real-time
• Use Ai actions: Explain, Simplify, Expand, Fix, Translate
• Browse version history and restore previous versions
• Copy or download your creation

Artifacts

When the Ai generates rich content (HTML apps, SVG graphics, interactive visualizations), it appears in the Artifacts Panel with a live preview. Toggle with /artifacts. You can interact with the content, view the source, or deploy it as a live web app.

Model Arena

Compare Ai models head-to-head with a blind comparison test:

1. Type /arena or go to the Arena via the command palette
2. Choose 2–3 models to compare
3. Send a prompt — all models respond simultaneously with anonymous labels ("Model A", "Model B")
4. Read the responses and vote for the best one
5. Model names are revealed after voting

Arena history tracks performance stats so you can see which models consistently perform best for your use cases.

Document RAG

Upload documents and let the Ai reference them in conversation:

1. Use the /docs command to open the document manager
2. Upload PDF, DOCX, or text files
3. Documents are automatically chunked and indexed
4. When you chat, relevant sections are injected as context — no extra steps needed
5. Source citations are included in Ai responses

Workflow Builder

Create automated multi-step Ai workflows with a visual drag-and-drop editor. Access with /workflow.

• Node types: Prompt, Condition (if/else), Tool, Transform, Input, Output, Loop, Delay
• Connect nodes by dragging between ports
• Execute workflows and watch nodes light up green/yellow/red in real-time
• Save workflows as reusable templates

App Deploy

Turn any HTML, JavaScript, or SVG creation from the Canvas or code blocks into a live web app. Access with /deploy.

1. Create or generate HTML/JS content in the Canvas
2. Click Deploy
3. Get a unique public URL instantly
4. Toggle between public and private access
5. Share with an embed snippet for websites

Agent Mode, Research & Planning

Agent Mode

Type /do followed by a task description to launch Agent Mode. The Ai creates a numbered step-by-step plan, then executes it autonomously — calling tools, reading files, searching the web, and writing code across up to 25 turns without waiting for you to respond.

The Agent Panel

When Agent Mode starts, a dedicated control panel appears showing:

• Goal — your original task description
• Plan — numbered steps with live progress indicators (active step highlighted, completed steps checked off)
• Status — current state: planning, executing, paused, waiting for approval, or complete
• Token usage — running count of input and output tokens consumed
• Execution log — color-coded log of every tool call, result, and decision

Controls

You stay in control throughout execution:

• Pause / Resume — pause execution at any point, then resume when ready
• Stop — end the run immediately and see a summary of what was accomplished

Risk-gated approvals

Tools are classified by risk level to balance autonomy with safety:

• Low risk (auto-approved) — read-only operations like file reads, web searches, and memory lookups
• Medium risk (auto-approved if you opt in) — file writes, code execution, and MCP tool calls
• High risk (always requires approval) — system commands, file deletion, and deployment actions

When a high-risk tool is triggered, the Agent Panel shows an approval dialog with the tool name and parameters. Click Approve to continue or Deny to skip that action.

User input requests

If the Ai needs a decision from you mid-task, an input field appears in the Agent Panel. Type your answer and the Ai continues from where it left off.

Parallel sub-agents

When a task involves multiple independent sub-tasks (e.g. researching 3 topics, comparing technologies), the Ai can spawn parallel sub-agents that work simultaneously. A sub-task panel appears in the Agent Panel showing each sub-agent's status with live progress updates. Results are collected and synthesized by the parent agent when all sub-tasks finish.

Background execution & auto-reconnect

Agent Mode is resilient to disconnections. If your browser tab closes, loses connection, or you navigate away, the agent keeps running on the server. When you return, CogmemUI automatically reconnects and replays any events you missed — you pick up right where you left off with no lost progress.

Run summary

When Agent Mode completes, you see a summary of total turns, tools used, files modified, and the outcome. Completed runs are also saved to your CogmemAi memory so you can recall what was done in future sessions.

Deep research mode

Type /research followed by a topic for thorough, multi-step research. The Ai will explore multiple angles, search the web for sources, cross-reference findings, and deliver a comprehensive, well-structured answer with evidence and citations.

Plan mode

Type /plan to toggle plan mode. In plan mode, the Ai outlines a structured approach before executing, letting you review and adjust the strategy before work begins. Useful for complex projects where you want to approve the approach first.

Mind Maps

Type /mindmap to visualize conversation topics, ideas, or any content as a spatial mind map. Nodes are interactive — click to expand, drag to rearrange, and let the Ai help you brainstorm by adding branches and connections.

Personas & Projects

Personas

Type /persona to manage Ai personas. Create custom personas with different system prompts, personalities, and tool configurations. Switch between personas to get different perspectives or specialized behavior for different tasks.

Projects

Type /project to organize your work into projects. Projects group conversations, memories, and documents together, keeping context separate between different workspaces.

Background tasks

Type /task to view and manage background tasks. Long-running operations like document processing, workflow execution, and scheduled jobs appear here with real-time status updates.

Scheduled Messages

Type /schedule to set up messages that send automatically at a specific time or on a recurring schedule. Use cases include daily briefings, regular data checks, scheduled reports, and timed reminders.

Git Integration

Type /git to open the Git panel. View repository status, diffs, branches, and commit history directly in CogmemUI. The Ai can also use git tools to help you manage code changes.

Minimap

Type /minimap to toggle a bird's-eye overview of long conversations. The minimap shows a scrollable thumbnail of the entire conversation, making it easy to navigate to any point in a lengthy thread.

Pinned Tools

The pinned toolbar appears above the input area for quick one-click access to your most-used tools. Right-click any tool to pin or unpin it. Pinned tools are saved to your browser and persist across sessions.

Available tools to pin:

@ Mentions

Type @ in the input to reference:

• @conversation:name — Pull in messages from another conversation as context
• @memory:subject — Inject a specific memory into the current prompt
• @model:name — Temporarily use a different model for this single message

Referenced content is automatically injected as context for the Ai — no copy-pasting needed.

Complete Slash Command Reference

Type / in the input to see all available commands. Here is the complete list:

Navigation & interface

Models & Ai

Panels & workspace

Features

Voice & output

Export & sharing

Ai Prefix Commands

Prefix commands modify your message before sending it to the Ai. Type the command followed by your text:

File & Directory Context

If your browser supports the File System Access API (Chrome, Edge), you can open a local folder and give the Ai context about your project files:

1. Open the Files Panel with /files or the panel button on the right
2. Click Choose Folder and select a directory
3. Browse the file tree and click files to view or include them
4. The Ai can reference your files when answering questions

Settings Reference

Open Settings with the gear icon in the header or type /settings. Here's what each tab does:

Keyboard Shortcuts

Ctrl + K

Open the command palette

Ctrl + N

Start a new conversation

Ctrl + F

Search within the current conversation

Ctrl + /

Show all keyboard shortcuts

Enter

Send message

Shift + Enter

New line (without sending)

Ctrl + Shift + C

Copy last Ai response

Escape

Close modals, panels, or menus

Teams

CogmemUI Teams turns your team’s scattered knowledge into shared Ai memory that proactively surfaces the right information to the right person at the right time. No searching. No asking. The Ai just knows — because your colleagues already taught it.

Teams require a CogmemAi Teams subscription ($49.99/seat/month). Learn more about CogmemAi Teams →

Creating a team

Go to Settings > Teams and click Create Team. Enter a team name and you’ll automatically become the team owner. You can create multiple teams — one for each project, department, or working group.

Inviting members

In the Teams settings tab, click Invite Member on your team. Enter the username of the person you want to add. They’ll receive an invitation they can accept or decline. You can also set their role when inviting:

• Owner — Full control. Can delete the team, manage all members, and change roles. One owner per team.
• Admin — Can invite and remove members, manage shared conversations, and access team memories.
• Member — Can view and participate in shared conversations and access team memories.

Switching to team mode

Click the team toggle button in the header (people icon) to activate your team. The button shows your active team name when enabled. If you belong to multiple teams, click the dropdown to switch between them instantly.

Sharing conversations

With team mode active, use the Share with Team option on any conversation to make it visible to all team members. Shared conversations appear in everyone’s sidebar with a team badge. You’ll see:

• Presence indicators — Colored avatars show who’s online right now
• Active viewers — See who’s viewing the same conversation as you
• Typing awareness — Real-time indicators when a teammate is typing in a shared conversation

Private conversations stay private. Only conversations you explicitly share become visible to your team.

Team memories — the knowledge base that seeks you

This is the core of CogmemAi Teams. When team mode is active, the Ai has access to team-scoped memories that are shared across every team member. As your team works, important knowledge automatically accumulates:

• Decisions and rationale (“We chose PostgreSQL because of ACID requirements”)
• Protocols and procedures (“Deploy to staging requires 2 reviewer approvals”)
• Known issues and fixes (“The payments endpoint has a race condition — use DB-level locking”)
• Preferences and conventions (“Use Zenodo for data repositories, not the institutional server”)

Proactive surfacing

Team memories don’t just sit in a database waiting to be searched. When any team member asks the Ai a question, relevant team memories are automatically retrieved and injected into the conversation. Ask “what’s our deployment process?” and the Ai already knows — because your colleague documented it last week.

Author attribution

Every piece of team knowledge includes the name of the person who contributed it. When the Ai surfaces a team memory, you see exactly who added it — so you can follow up with the right person if you need more detail.

Scope isolation

Personal memories and team memories are completely separate. Your personal memories stay yours — they are never visible to teammates. Team memories belong to the team. Toggle between personal and team memory scope using the scope selector in the Memory Panel. When team mode is off, the Ai only uses your personal memories. Zero bleed between scopes.

Multiple teams

You can belong to as many teams as you need. Each team has its own shared memory, shared conversations, and member list. Switch between teams with one click in the header dropdown. Common setups:

• A team per project or client
• A team per department (engineering, design, marketing)
• A team per research group or lab
• A team per shift or rotation (healthcare, support)

Managing your team

Team owners and admins can manage the team from Settings > Teams:

• Change roles — Promote members to admin or demote admins to member
• Remove members — Remove someone from the team (their personal data is unaffected)
• Delete team — Owner only. Permanently removes the team and all shared memories

Extensions

The Extension API lets you build and install custom functionality:

• Custom slash commands — Add your own /commands
• Custom tools — Give the Ai new capabilities
• UI panels — Add custom interface elements
• System prompt fragments — Automatically inject instructions
• Event hooks — React to chat events with custom logic

Install extensions via file upload or HTTPS URL in Settings > Extensions. All URL-based installs show a code preview for security review before execution.