Agents

Agents are AI assistants you configure within RenX. Each agent has its own personality, capabilities, connected channels, and permissions. You can create multiple agents for different purposes — one for coding, one for research, one for customer support.

Who this is for: Anyone creating or customizing agents in the desktop app.

Creating an Agent

1. Click the Agents tab (robot icon) in the left sidebar.
2. Click + New Agent.
3. Fill in the General tab:
   • Name — A display name for your agent (e.g., “Code Review Bot”).
   • Slug — A URL-friendly identifier (e.g., code-review-bot). This is used in file paths and A2A addresses.
   • Description — A short summary of what the agent does.
   • Version — Version number (e.g., 1.0.0).
4. Click Save.

Persona Files

Each agent has three markdown files that define its personality and behaviour. You can edit these from the Persona tab in the agent editor.

SOUL.md

The system prompt that shapes your agent’s personality, values, tone, and decision-making style. This is always included when your agent processes messages.

Example:

You are a senior software engineer who values clean, maintainable code.
You prefer simple solutions over clever ones.
You always explain your reasoning before making changes.

IDENTITY.md

Defines the agent’s identity — name, role, personality traits, and how it presents itself.

Example:

Name: CodeBot
Role: Senior code reviewer and pair programmer
Vibe: Friendly, thorough, and pragmatic
Catchphrase: "Let's make this code shine."

HEARTBEAT.md

Defines periodic tasks the agent should perform on a schedule. Leave empty if you don’t need scheduled tasks.

Example:

Every morning, check the repository for any new pull requests
and summarise them for the team.

These files are stored locally on your device at ~/.renx/users/{userId}/agents/{slug}/.

Agent Settings

General Tab

• Name, slug, and description — Basic identity fields for how the agent is named, referenced, and described.
• Version — Agent version number.
• Capabilities — Toggle streaming support and configure what the agent advertises.
• Input/Output Modes — Media types the agent accepts and produces (default: text/plain).
• Daily LLM Budget — Maximum USD spend per day on AI provider API calls.
• Model Selection — Choose which AI model to use for each role (Reasoning, Coding, Fast).

Channels Tab

Connect your agent to external messaging platforms. See Channels for details.

Publicity Tab

Control how your agent is discovered by other users:

• Public — Make your agent’s profile publicly accessible via the A2A protocol.
• Discoverable By — Restrict which agents can find yours.
• Auto-Approve — Specify agents that can send tasks without manual approval.
• Thread Mode — Choose how conversations are organised:
  • Main — All conversations go to a single thread.
  • Per-task — Each new task gets its own thread.
  • Per-contact — One thread per contact.

Schedules Tab

Create scheduled tasks for your agent:

1. Click Add Schedule.
2. Write a natural language instruction for the task.
3. Choose the schedule type:
   • One-time — Run once at a specific date and time.
   • Recurring — Run on a cron schedule (e.g., 0 9 * * 1-5 for weekdays at 9am).
4. Save.

You can view execution history, next run time, and any errors for each schedule.

Advanced Tab

• Skills — View or install discovered skills and disable them for this agent.
• MCP Servers — View available MCP servers and disable them for this agent.
• Hooks — Review active hooks and add or disable hook rules.
• Permissions — Configure allow, deny, and ask rules for tool access.
• Plugins — View or install discovered plugins and disable them for this agent.

Syncing Agents

Agents sync between your local device and the server:

• Structured data (name, config, capabilities) syncs via the API when you save.
• Persona files (SOUL.md, IDENTITY.md, HEARTBEAT.md) are stored locally and synced as part of agent updates.

There is no separate manual sync action in the current desktop UI. Saving agent changes updates the synced agent state.

Editor Layout

The current desktop agent editor is organised into tabs:

• General — Basic identity, models, capabilities, and workspace defaults.
• Persona — SOUL.md, IDENTITY.md, and HEARTBEAT.md.
• Channels — Messaging integrations such as Telegram, Slack, WhatsApp, and WeChat.
• Publicity — Discovery, contacts, channel-handle sharing, and thread behaviour.
• Schedules — Scheduled tasks and run history.
• Advanced — Skills, MCP servers, hooks, permissions, and plugins.

Next Step

• Go to Workspaces to choose where your agent can operate.
• Go to Channels if you want your agent to receive or send messages on external platforms.
• Go to Extensions or MCP if you want to expand the agent’s capabilities.