Documentation
Learn the complete OpenLioness operating system.
This guide explains how to create agents, connect social channels and website widgets, add skills, build workflows, manage leads and sessions, monitor instances and usage, run cronjobs, consolidate memory, inspect logs, and configure settings.
Quick path
Settings → Agents → Knowledge → Skills → Channels → Sessions → Workflows.
Start here
Getting started
Understand how OpenLioness is structured before creating your first AI agent.
OpenLioness is one AI operating workspace for agents, skills, workflows, knowledge, channels, sessions, leads, instances, usage, cronjobs, dreaming, logs, and settings.
Start by going to Settings and saving your model provider API key and model name. After that, go to Agents and create your first agent.
Once your agent works, connect knowledge, create skills, attach a channel such as Website Widget or WhatsApp, and then build workflows or cronjobs for repeated work.
Checklist
Login or create an OpenLioness account.
Open Settings and add provider API key plus model name.
Create one agent from the Agents page.
Test the agent in Sessions before connecting automations.
Build
Agents
Agents are AI workers that can answer, qualify, research, operate workflows, and handle channel conversations.
Before creating agents, save your provider key and model configuration in Settings. This keeps agent setup clean and avoids hardcoded model details.
After provider setup, create agents with name, description, capability, model, status, and instructions.
Agents can be connected to social channels, website widgets, knowledge sources, skills, workflows, and sessions.
Checklist
Save provider API key in Settings.
Save model name and provider metadata.
Create an agent with a clear role and prompt.
Assign knowledge, skills, channels, or workflows after testing.
Capabilities
Skills
Skills are reusable actions that agents and workflows can run.
Create skills for browser automation, website research, API calls, webhooks, file work, database saves, and other repeatable operations.
Each skill should have a clear type, safe configuration, input example, and status.
Test skills manually first, then connect them to agents or workflows once the output is stable.
Checklist
Create a skill from the Skills page.
Choose the skill type such as Browser Automation or API.
Add description and sample run input.
Run a manual test before using it in workflows.
Automation
Workflows
Workflows connect agents, skills, inputs, and outputs into multi-step automations.
Use workflows for repeatable tasks such as browser-to-file-to-database, lead processing, article posting, research summaries, and operational actions.
Workflows can be saved as templates, edited with JSON, run manually, or triggered from cronjobs.
Always check the latest result, duration, step count, and status after a workflow run.
Checklist
Create a workflow with a clear name and description.
Define each step in simple language or JSON.
Run once manually and inspect output.
Use cronjobs only after the workflow is stable.
Memory
Knowledge
Knowledge sources give agents trusted business and product context.
Add FAQs, product details, policies, website notes, file text, and business rules from the Knowledge page.
Knowledge can be workspace-wide or scoped to one specific agent.
Better knowledge means better agent answers. Keep sources short, clean, and updated.
Checklist
Create a knowledge source with a clear title.
Choose source type such as Text or file-based content.
Scope it to workspace-wide or a specific agent.
Update outdated knowledge when agent answers become weak.
Surfaces
Channels
Channels connect agents to social apps, website widgets, WhatsApp, email, and APIs.
Channels are where conversations enter OpenLioness. You can connect social-style surfaces such as Facebook and Instagram, WhatsApp, Telegram, Slack, Email, Public API, and Website Widget.
A social account or WhatsApp channel can be assigned to an AI agent so incoming conversations become agent-powered sessions.
For a website widget, create the channel, select the default agent, copy the widget script, and paste it on your website.
Checklist
Choose a channel type such as Website Widget, WhatsApp, Email, or Public API.
Name the channel clearly.
Assign a default agent or select later.
Copy the website widget script and install it on your site when needed.
Conversations
Sessions
Sessions store saved conversations from website widgets, agent chats, and connected channels.
Use Sessions to review what visitors asked, which agent replied, and when activity happened.
Sessions can come from website widget conversations, social channels, WhatsApp, or internal agent chat.
Review sessions to discover repeated questions, missing knowledge, and opportunities for workflows or lead follow-up.
Checklist
Filter sessions by Website or Agent Chat.
Open important conversations for review.
Convert repeated questions into Knowledge content.
Turn business-ready conversations into Leads.
Capture
Leads
Leads store captured visitor details, source context, company data, and follow-up status.
Leads are created when visitor information or business context should be saved for follow-up.
You can review contact details, company, source, context, status, and updated time.
Use filters and export CSV when you need to send leads to sales, support, or another system.
Checklist
Review captured name, email, phone, company, and source.
Filter by status or source.
Open lead context before follow-up.
Export CSV when needed.
Runtime
Instances
Instances monitor connected runtimes, widgets, workflow runners, and channel services.
Use Instances to register and monitor runtime health for website widgets, workflow runners, browser workers, or channel services.
Each instance can show type, status, endpoint or handle, and version details.
If a channel or automation is not responding, Instances is one of the first places to check.
Checklist
Register the runtime name and type.
Set status and endpoint or handle.
Monitor online, offline, warning, and error states.
Refresh runtime health after deployment changes.
Analytics
Usage
Usage shows requests, token volume, latency, provider breakdown, model breakdown, and daily usage.
Usage helps you understand how much work your agents are doing and how providers are performing.
Track total requests, total tokens, average tokens, average latency, failed events, provider usage, model usage, and daily trends.
Use this page to spot expensive agents, slow model responses, and failed events.
Checklist
Review total requests and tokens.
Check average latency and success rate.
Compare provider and model breakdown.
Investigate failed events from logs.
Schedule
Cronjobs
Cronjobs schedule recurring agent prompts, reports, summaries, scraping, reminders, and background intelligence.
Cronjobs are useful when you want an agent or workflow to run on a schedule without manual clicks.
Use presets such as daily reports, evening summaries, weekly kickoff, or custom cron expressions.
Always review latest run details and pause jobs that are failing repeatedly.
Checklist
Choose agent or workflow to run.
Set the schedule preset or cron expression.
Run once or review first scheduled output.
Pause, resume, or delete when needed.
Memory consolidation
Dreaming
Dreaming stages useful session facts, groups recurring patterns, and promotes high-value memories.
Dreaming is a memory consolidation layer for agents. It can scan recent sessions, identify useful recurring facts, and promote only important memories.
Use it to reduce noise and keep long-term memory useful for agent context.
Run sweeps by lookback period, session count, candidate count, and minimum score.
Checklist
Choose lookback period such as last 24h or 7 days.
Set session and candidate limits.
Run a sweep and inspect latest run.
Promote only useful memories.
Observability
Logs
Logs show workspace activity, automation errors, debugging events, and system actions.
Logs are the debugging layer for OpenLioness. Use them when workflows, instances, skills, or channels behave unexpectedly.
Filter by message, source, entity, level, debug, info, warning, or error.
Open metadata preview when you need more detail about a record.
Checklist
Filter by source or entity type.
Check warnings and errors first.
Open metadata preview for details.
Use logs together with Usage and Instances.
Admin
Settings
Settings manage identity, workspace branding, provider keys, model names, and default controls.
Settings is where you save provider API keys before creating useful agents.
You can update profile details, workspace name, workspace slug, brand name, logo URL, accent color, widget defaults, runtime defaults, dreaming defaults, and system preferences.
Provider keys are saved as encrypted metadata and should never be hardcoded in frontend files.
Checklist
Save provider API key and label.
Save model provider and model name.
Update brand name, logo URL, and accent color.
Review workspace controls before production use.
Fix issues
Troubleshooting
Use this section when agents, skills, channels, or workflows are not behaving as expected.
If an agent cannot answer, check Settings for provider key and model name, then inspect the agent prompt and active knowledge.
If a website widget does not appear, confirm the channel is created, the correct agent is selected, and the script is copied into the website.
If a social or WhatsApp channel is not responding, check channel status, instances, sessions, logs, and provider availability.
Checklist
Check provider keys in Settings.
Check channel status and assigned agent.
Review Logs for warnings or errors.
Restart dev server after changing globals.css or theme variables.