Dashboard

Lex

← Agents
Identity
@
Handle used in routing
L
Short label shown in greetings and UI
Injected into the system prompt as who this agent is. Used in both modes.
Tone & Personality
Added to system prompt. Applies to all responses.
Behavioral instructions for how the agent thinks and communicates. Keep specific.
Purpose
Why this agent exists. Anchors every decision Claude makes.
What success looks like. Tells Claude how to evaluate its own output.
Tasks

What this agent can do. Injected into the system prompt as RESPONSIBILITIES. Each task maps to a skill or capability.

Constraints & Context
Rules this agent must never break. Injected into system prompt. Both modes. Manage platform-wide rules →
Always-on situational context: user's business, priorities, habits.
Permissions
In agentic mode, these permissions control which tools Claude can call during reasoning.
In classic mode, they control which integration data gets fetched into the prompt.
Access Airtable
Read and write Airtable bases, tables, and records
Requires Airtable
Access Calendar
View upcoming events and daily schedule
Requires Google
Build Website
Create websites and deployable files
Requires Artifact
Access Email
Read inbox for action items and urgent messages
Requires Google
Access CRM
Read contacts, opportunities, and form submissions from GoHighLevel.
Gohighlevel
Access Analytics
Read Google Analytics 4 traffic and session data
Requires Google_analytics
Access Search Console
Read Google Search Console query and ranking data
Requires Google_search_console
Access Google Sheets
Write keyword research results to Google Sheets.
Requires Google
Access Keywords Everywhere
Fetch search volume, CPC, and competition data for keyword lists.
Requires Keywords_everywhere
Access Meetings
Review meeting transcripts and action items
otter (soon)
Access Projects
Check task deadlines and project status
Asana
Access Search (SerpAPI)
Run Google searches via SerpAPI for SERP data and keyword expansion.
Requires Serpapi
Deploy to Vercel
Deploy websites and static pages to Vercel for a live URL
Requires Vercel

Skills this agent can perform. Users invoke them in Telegram with a slash command, e.g. /brief. Use {google_calendar}, {asana}, {gohighlevel} etc. in prompts to inject live data.

+ New Skill
/review
Document Review
On Edit
/legal
Legal Email Review
On Edit
Team Membership

Teams this agent belongs to. An agent can be a member of multiple teams with different roles.

Team Role
Department Heads Member
Primary Team (Legacy)
When set, this agent becomes a sub-agent routed through its primary team leader.
Visual Identity
Circle fill color in the mobile app
Used for pronoun copy in the app
Voice
Voice used for TTS in the mobile app. Leave blank to disable voice for this agent.
Default Expression
Default facial expression shown in the idle animation
Enable user knowledge base
When enabled, users can upload their own documents to personalise this agent.
Upload Files
Drop files here or click to browse
Supported: PDF, DOCX, TXT, Markdown, MP3, M4A, WAV, OGG, FLAC - max 25 MB per file
YouTube / Podcasts
Paste a YouTube or podcast URL and the agent will transcribe and learn from it.
  • YouTube video: open the video, copy the URL from the address bar (e.g. youtube.com/watch?v=abc123)
  • YouTube channel: go to the channel page and copy its URL (e.g. youtube.com/@channelname)
  • Podcast RSS feed: search for your podcast on podcastindex.org or castos.com, then copy the RSS link
Paste Text
Articles, writing samples, SOPs, personality descriptions, anything text-based.
Loading knowledge items...
Model
Fast model (Haiku) used to classify intent and pick the right agent.
Model used to generate responses. In agentic mode, called on every loop iteration.
Agentic Mode
ON: Claude reasons over the user's message, decides which tools to call, and loops until the task is complete. Skills use a Goal field instead of a Prompt.
OFF (Classic): Claude receives a pre-assembled prompt with integration data already injected. Skills use a Prompt template with {tokens}.
Routing
Enable Routing
Whether Haiku classifies this agent's messages before execution.
Email

Users configure their agent email and response email on their individual agent settings page.

Native Routing
Routes all messages through the agent module directly. Irrelevant when agentic is ON.
Security
Primary Agent
Default agent for new conversations. Only one agent can be primary — saving will unset it on all others.
Admin Agent
Grants this agent write access to agents, skills, and permissions tables.
Changes saved. Restart the agent bot to apply.   Go to Dashboard →