Bot Management

Prerequisites: Before reading this guide, it’s recommended to familiarize yourself with the resource model and permission model sections in Core Concepts.

Bots are the managed AI agents within the Monstrum platform. This guide covers everything about creating, configuring, resource binding, channel integration, and lifecycle management of Bots.

Overview

Bots are the core entities of the Monstrum platform. Each Bot has its own:

  • Identity — Name, description, and workspace membership
  • Permissions — Controlled through role assignment and resource binding to determine available tools and scope of operations
  • Resource Bindings — Determine which external systems the Bot can access (SSH servers, MCP tools, other Bots, etc.)
  • Credential Isolation — Bots never have access to plaintext credentials; the platform decrypts and injects them at execution time
  • Memory — A persistent knowledge base across sessions, supporting both automatic extraction and manual management
  • Budget — Monthly Token consumption cap; execution automatically stops when the budget is exceeded

Bots follow the platform’s core security principles: Least Privilege (unauthorized tools are completely invisible to the Bot), Credential Isolation (LLM never has access to secrets throughout the entire process), and Fail-Closed (if anything goes wrong, the Bot can do nothing rather than operate unrestricted).

Creating a Bot

Click Bot Management in the left navigation bar to access the Bot list page.

Creation Steps

  1. Click the Create Bot button in the upper right corner
  2. Fill in basic information:
    • Bot Name — Required, must be unique within the workspace. Use descriptive names that reflect the Bot’s purpose, such as “Ops Inspector” or “Customer Support Assistant”
    • Description — Optional, describes the Bot’s role and responsibilities. This description is injected into the system prompt to help the LLM understand its role
  3. Click Create

After creation, the Bot defaults to “Running” status using the workspace’s default LLM provider and model. You can immediately start chatting with it on the Overview page, or configure it in detail on the Settings page first.

If the workspace has no LLM provider configured, the page will display a “No LLM provider configured” prompt, and you’ll need to add a provider on the LLM Providers page first.

Bot Detail Page

Click a Bot card or its name in the list to enter the Bot detail page. The detail page organizes different functional areas through tabs:

TabDescription
OverviewBot status overview + built-in chat interface
Resource ConfigManage the Bot’s bound Resources and permissions
ChannelsConfigure Slack, Feishu, Telegram, and other IM integrations
Execution HistoryView historical task execution records and details
Scheduled TasksConfigure periodic automated tasks
WorkflowsManage workflows associated with the Bot and their binding modes
MemoryView and manage the Bot’s persistent memory
PromptsCustomize Bot-level prompt templates
SettingsAll Bot configurations including identity, model, constraints, and capabilities

Overview Tab

The Overview Tab is the default view when entering the Bot detail page, containing two main areas.

Status Panel

The top of the page displays the Bot’s key operational metrics:

  • Current Status — Running / Paused / Disabled
  • Monthly Cost — Current month’s Token consumption converted to USD, with daily average cost
  • Call Efficiency — Number of LLM calls and resource calls
  • Permission Pass Rate — Percentage of tool calls that passed permission checks, along with the number of rejections
  • Token Consumption — Cumulative Token count consumed
  • Model — Current LLM model in use
  • Quick Actions — Create task, view task history

Chat Interface

The bottom of the page embeds a full chat interface where you can interact directly with the Bot:

  • Type a message in the input box and press Enter to send
  • The Bot uses the configured LLM for reasoning and replies, with streaming output support
  • If the Bot invokes a tool, the interface displays the tool call name, parameters, and execution results
  • Click New Conversation to start a fresh session, clearing previous conversation context
  • Click End Session to manually close the current session, triggering memory extraction (if auto memory is enabled)

For detailed information about the chat functionality, see Chat Interaction.

Settings Tab

The Settings Tab is the Bot’s core configuration page, containing multiple configuration areas including identity & behavior, model configuration, execution & constraints, and agent capabilities.

Identity & Behavior

System Prompt

The system prompt defines the Bot’s identity awareness and behavioral patterns.

  • Custom Prompt — Enter the instructions you want the Bot to follow in the text box. For example: “You are a professional operations engineer skilled in server monitoring and troubleshooting. Confirm with the user before executing any dangerous operations.”
  • Leave Empty for Default — When left blank, the Bot uses the workspace-level default template (or the platform’s built-in default template if the workspace hasn’t customized one either)

Important: Custom system prompts only replace the base template portion. The following content is always automatically appended to the end of the system prompt and does not need to be added manually:

  • Available Resource Summary — Tool descriptions auto-generated based on the Bot’s bound Resources
  • Bot Memory — Relevant memory entries loaded from the memory store
  • Skill Instructions — Detailed operational instructions for enabled skills
  • Reasoning Mode Instructions — Framework instructions for planning or adaptive modes

Bot Name & Description

You can also modify the Bot’s name and description in the settings. The description content is injected into the system prompt, so it’s recommended to summarize the Bot’s responsibilities in one or two sentences.

Model Configuration

LLM Provider

By default, the Bot uses the workspace’s default LLM provider. You can select a different provider for a specific Bot:

  1. Choose a configured provider from the “LLM Provider” dropdown
  2. If you need to create a new provider, click New Provider to jump directly to the creation page

Select “Use workspace default provider” to revert to the default.

Model Selection

After selecting a provider, the model dropdown displays all models supported by that provider. Choose a model that fits your use case:

  • Complex reasoning and multi-step tool calls: Use a more capable model
  • Simple conversations and classification: Use a lighter model to reduce costs

Temperature

Temperature controls the randomness of LLM output:

  • 0 — Most deterministic output, suitable for scenarios requiring stable and reproducible results (e.g., operations tasks)
  • 0.5-0.7 — Balances creativity and consistency, suitable for most scenarios
  • 1 — Most diverse output, suitable for creative writing and similar scenarios

Execution & Constraints

Max Iterations

Controls the upper limit of reasoning-tool call loops in a single task. Each loop consists of: LLM reasoning -> tool call -> get result -> reason again.

  • The default value is sufficient for most scenarios
  • If the Bot needs to handle complex multi-step operations (e.g., executing multiple commands in sequence), increase this value appropriately
  • When the limit is reached, the Bot stops execution and indicates “Maximum execution rounds reached” in its reply

Timeout

The timeout in seconds for a single task. The task is forcefully terminated after this duration.

  • For simple conversation scenarios, the default value is sufficient
  • For tasks involving long-running tool calls (e.g., large file transfers, complex queries), increase this value appropriately
  • After timeout, the Bot indicates “Request timed out” in its reply

Monthly Budget

Set the Bot’s monthly Token consumption cap in USD.

  • Budget Warning — When monthly consumption reaches 80% of the budget, the platform sends a budget warning notification
  • Auto Stop — When monthly consumption reaches 100%, the Bot automatically stops executing new tasks until the next month’s budget resets
  • Leave Empty — No budget cap (not recommended for production environments)

You can view the current monthly cost and daily average consumption in the Overview Tab’s status panel.

Agent Capabilities

Reasoning Mode

The reasoning mode determines how the Bot processes tasks:

ModeBehaviorUse Cases
ReactiveReasons and executes directly upon receiving a message, without upfront planningSimple conversations, single-step operations
PlanningOutputs an execution plan first, then executes step by step according to the planMulti-step operations, scenarios requiring predictable execution paths
AdaptiveCreates a plan first, then reflects and adjusts the plan based on results during executionComplex tasks, scenarios with uncertain outcomes
  • Reactive is the default mode and consumes the fewest Tokens
  • Planning and adaptive modes consume additional Tokens (for generating and maintaining plans) but provide better results for complex tasks

Auto Memory Extraction

When enabled, the Bot automatically extracts valuable information from conversations and saves it to the memory store at the following moments:

  • Task Completion — After a task finishes executing in Task mode
  • Session Expiration — When a session automatically closes due to idle timeout in Session mode

Extracted memories are automatically categorized (user preferences, rules, events, knowledge, etc.) and assessed for importance. In the next conversation, relevant memories are injected into the system prompt, helping the Bot “remember” previous interactions.

For detailed information about the memory system, see the “Memory Management” section in this document.

Self-Created Scheduled Tasks

When enabled, the Bot can create scheduled tasks on its own during conversations. For example, you can tell the Bot “Check server status every day at 9 AM,” and the Bot will automatically create the corresponding scheduled task.

When this option is disabled, scheduled tasks can only be created manually by users in the Scheduled Tasks Tab.

Skills

Skills are predefined instruction sets that inject domain-specific operational knowledge into the Bot.

  1. Check the skills you want to enable in the skill selector
  2. Selected skill instructions are appended to the Bot’s system prompt
  3. Skills are sourced from the Skills management page (left navigation bar Skills) and support uploading custom skill packages

Resource Config Tab

The Resource Config Tab manages the binding relationship between the Bot and Resources. A Bot can only use tools provided by bound Resources — unbound Resources are completely invisible to the Bot.

Binding Resources

  1. Click the Bind Resource button
  2. Select the Resource to bind from the dropdown (you need to create it first on the Resource Management page)
  3. Choose a credential (if the Resource has multiple Credentials, select one; if there’s only one, it’s auto-selected)
  4. Configure permissions:
    • Static tool types (SSH, Bot, Web3, etc.): Select allowed operations, such as ssh_execute, balance.get, etc. Leave empty to allow all operations
    • Dynamic tool types (MCP, Monstrum Agent): Display the list of discovered tools, check the ones the Bot is allowed to use
  5. Click Confirm to complete the binding

Binding via Roles

In addition to directly binding Resources, you can batch-grant permissions through Bot Roles:

  1. In the “Bot Roles” area of the resource config page, click Assign Role
  2. Select a defined role
  3. All resource bindings defined by the role take effect automatically

Roles are collections of predefined resource bindings and permission configurations, suitable for scenarios where multiple Bots need the same permission policies. See Roles & Permissions for details.

Editing Permissions

After binding, you can click Edit Permissions on any resource entry at any time to modify the allowed scope of operations. Changes take effect immediately without restarting the Bot session.

Unbinding Resources

Click the Unbind button on a resource entry to remove the binding. After unbinding, the Bot will no longer see that Resource’s tools in its next reasoning cycle.

If there are no resources available to bind, the page will prompt you to go to the Resource Management page to create one. See Resources & Credentials for details.

Channels Tab

The Channels Tab manages the Bot’s external IM integrations. Once configured, the Bot can receive and reply to messages on platforms like Slack, Feishu, Telegram, and Discord.

Adding a Channel

  1. Click Add Gateway
  2. Select the channel type:
    • Webhook — Generic HTTP callback, suitable for custom integrations
    • Slack — Requires Signing Secret and Bot Token
    • Feishu — Requires App ID and App Secret
    • Telegram — Requires Bot Token
    • Discord — Requires Bot Token
  3. Fill in the corresponding authentication information
  4. Configure the callback URL (Bot replies are sent to this address)
  5. Save

Channel Management

After creation, you can:

  • Enable/Disable — Temporarily deactivate a channel without deleting its configuration
  • Test Connection — Verify that the channel configuration is correct
  • View Webhook URL — Copy this URL to the corresponding platform’s Webhook configuration
  • Delete — Permanently remove the channel configuration

Each channel is bound to one Bot. Messages from that channel are automatically routed to the bound Bot for processing.

See Gateway for details.

Execution History Tab

The Execution History Tab displays all task execution records for the Bot.

Task List

Each record in the list contains:

  • Task ID — Unique identifier
  • Instruction — The message content that triggered the task
  • Status — Pending / Running / Completed / Failed / Cancelled
  • Tokens — Number of Tokens consumed by the task
  • Created At — When the task was created

Filtering & Search

  • Use the search box for fuzzy search by instruction content
  • Use the status filter to filter by task status

Task Actions

  • Details — View the complete execution timeline of the task, including detailed records of each LLM request, tool call, and permission check
  • Retry — Re-execute a failed task
  • Cancel — Cancel a running or pending task

Execution Timeline

Click “Details” to enter the task’s execution timeline view, where you can see the complete execution process:

  1. Task Start — Records the trigger source and initial instruction
  2. LLM Request/Response — Input and output of each reasoning round
  3. Tool Call — Which tool was called and what parameters were passed
  4. Tool Result — The return value of the tool execution
  5. Permission Check — Whether each tool call passed Guardian’s permission verification, and the reason when rejected
  6. Task Completion/Failure — Final status and result

Scheduled Tasks Tab

The Scheduled Tasks Tab manages the Bot’s periodic automated tasks. Once configured, the Bot automatically executes specified operations on schedule.

Creating a Scheduled Task

  1. Click Add Scheduled Task
  2. Fill in the configuration:
    • Task Name — e.g., “Daily Inspection”, “Weekly Report Generation”
    • Schedule — Select frequency (hourly/daily/weekly/monthly) and specific time, or switch to advanced mode to use Cron expressions
    • Timezone — Select the execution timezone (e.g., Asia/Shanghai, UTC)
    • Task Instruction — Tell the Bot what to do when triggered, e.g., “Check server status and generate a report”
    • Target Type — Choose “Send Task” (direct Bot execution) or “Run Workflow” (trigger a specified workflow)
    • Enable Immediately — Whether to activate immediately after creation
  3. Save

Schedule Configuration

The platform provides two configuration methods:

Simple Mode:

  • Hourly — Set which minute of each hour to execute
  • Daily — Set execution time (hour:minute)
  • Weekly — Select day of the week + execution time
  • Monthly — Select day of the month + execution time

Advanced Mode (Cron):

Switch to advanced mode to enter standard Cron expressions directly. For example:

  • 0 9 * * * — Every day at 9:00
  • 0 9 * * 1-5 — Weekdays at 9:00
  • */30 * * * * — Every 30 minutes

After configuration, the page previews the upcoming execution times.

Managing Scheduled Tasks

  • Enable/Disable — Temporarily pause or resume scheduled tasks
  • Edit — Modify task name, schedule, or instructions
  • Delete — Permanently remove
  • Execution History — View historical trigger records and the status of each execution

Each scheduled task record displays the last execution time and next execution time for easy monitoring.

Workflows Tab

The Workflows Tab manages workflows associated with the current Bot. Workflows are visual multi-step automation processes that can orchestrate multiple steps, conditional branches, and parallel execution.

Workflow Binding Modes

The association between a Bot and workflows is controlled through “binding modes,” which determine how workflows are triggered:

Binding ModeDescription
AI AutoThe Bot automatically determines whether to trigger the workflow based on user intent during conversation
CommandUsers trigger via a specific command name, e.g., /deploy
Auto + CommandBoth methods can trigger the workflow
UnboundThe workflow is not available to the Bot

Configuring Bindings

  1. Navigate to the Workflows Tab, which lists all activated workflows
  2. Select a binding mode for each workflow
  3. If you choose “Command” or “Auto + Command” mode, enter the command name (e.g., deploy)
  4. Click Save to apply the binding

If the workspace has no activated workflows, the page will prompt you to create and activate workflows on the Workflows page first.

Usage

  • AI Auto Mode: Describe your needs naturally in conversation, and the Bot will automatically determine whether to use a workflow
  • Command Mode: Trigger by entering a command in conversation, e.g., “Run the deploy workflow”

For creating and editing workflows, see Workflow Automation.

Memory Management

The Memory Tab manages the Bot’s persistent knowledge base. Memories are retained across sessions, helping the Bot “remember” important information.

Memory Sources

Bot memories come from two sources:

  • Auto Extraction (source=auto) — When “Auto Memory Extraction” is enabled, the Bot automatically extracts information from conversations when tasks complete or sessions expire
  • Manual Addition (source=user) — Entries manually created by users in the Memory Tab

Auto-extracted memories are fully replaced during the next extraction (within the same scope), while manually added memories are always retained and never overwritten automatically.

Memory Partitions

Memories are stored in partitions by scope:

ScopeDescriptionInjection Timing
GlobalGeneral memories without a specific scopeAlways injected
ChannelMemories associated with a specific Gateway channelInjected during corresponding channel sessions
TaskMemories associated with a specific taskInjected during corresponding tasks
ResourceMemories associated with a specific ResourceInjected when using the corresponding resource
Scheduled Task / WorkflowMemories associated with a specific scheduled task or workflowInjected in the corresponding context

Viewing Memories

The Memory Tab displays all memory entries with support for:

  • Search — Search memory content by keywords
  • Browse by Category — Filter by category (user preferences, rules, events, knowledge)

Each entry displays its content, category, and importance level (1-10).

Manually Adding Memories

  1. Click Add Entry
  2. Select a category:
    • User Preferences — User habits and preference settings
    • Learned Rules — Rules and conventions the Bot should follow
    • Important Events — Key events that need to be remembered
    • Knowledge Base — Domain-specific knowledge and information
  3. Enter the memory content
  4. Set the importance level (1-10, higher numbers mean higher priority)
  5. Save

Editing & Clearing

  • Edit — Click an entry to enter edit mode, modify the content, and save
  • Clear — Delete all memory entries (irreversible, requires confirmation)

Prompts Tab

The Prompts Tab allows you to customize Bot-level prompt templates. The platform uses multiple prompt templates to control Bot behavior in different scenarios.

Three-Tier Priority

Prompt templates have three priority levels, from highest to lowest:

  1. Bot Custom — Templates set in this Tab
  2. Workspace Template — Default templates defined in workspace settings
  3. System Default — Platform built-in default templates

Each template item is labeled with its current source (“Bot Custom”, “Workspace”, or “System Default”).

Configurable Templates

TemplatePurpose
Default Task PromptBase system prompt in Task mode
Default Session PromptBase system prompt in Session mode
Group Chat PromptAdditional instructions for group chat scenarios
Planning Reasoning PromptFramework instructions for planning mode
Adaptive Reasoning PromptFramework instructions for adaptive mode
Memory Extraction PromptControls auto memory extraction behavior
Conversation Summary PromptControls long conversation compression and summarization behavior

Customizing Templates

  1. Click the template item you want to modify
  2. Edit the content
  3. Save

Click Reset to Workspace Default to clear the Bot-level customization and revert to workspace or system default templates.

In most scenarios, you don’t need to modify prompt templates — the defaults are already optimized. Only adjust them when you have a specific need, such as completely changing the Bot’s reasoning style or conversation strategy.

Bot Cloning

When you need to create a Bot with similar configuration, you can use the clone feature.

Steps

  1. On the Bot list page, find the Bot you want to clone
  2. Click the Clone button on the Bot card
  3. Confirm the clone operation

Clone Scope

Cloning copies the following:

  • Bot name (automatically appended with a suffix to avoid duplicates)
  • Description
  • System prompt
  • Model configuration (provider, model, Temperature)
  • Execution constraints (max iterations, timeout, monthly budget)
  • Agent capability settings (reasoning mode, auto memory, self-created scheduled tasks)
  • Skill configuration
  • Resource bindings and permissions

Cloning does not copy the following:

  • Conversation history
  • Memory entries
  • Execution history
  • Scheduled tasks
  • Channel configurations

Bot Groups

Bot grouping helps you organize and categorize Bots when you have a large number of them.

Creating a Group

  1. On the Bot Management page, click Create Group
  2. Enter the group name
  3. Confirm

Managing Groups

  • Drag or add Bots into groups
  • Each Bot can belong to one group
  • Ungrouped Bots appear in the default list
  • Groups can be renamed or deleted

Groups are for organizational purposes only in the UI and do not affect Bot permissions or behavior.

Bot Call Graph

When multiple Bots in a workspace call each other through Bot-type Resources, the call graph helps you visualize these relationships.

How to View

On the Bot Management page, click the Call Graph button to display the calling relationships between all Bots in the workspace as a DAG (Directed Acyclic Graph).

Graph Content

  • Nodes — Each Bot is displayed as a node with its name and description
  • Edges — Point from the caller to the callee, indicating that one Bot has bound another Bot as a resource
  • Current Bot — If entered from a Bot detail page, the current Bot is highlighted

Typical Scenarios

The call graph is particularly useful for:

  • Troubleshooting permission issues in Bot delegation chains
  • Understanding the call topology of orchestrator Bots
  • Identifying redundant or circular call relationships

Pausing & Resuming

You can temporarily pause a Bot without deleting any of its configuration.

Pausing a Bot

  1. On the Bot list page, click the Pause button on the Bot card
  2. The Bot status changes to “Paused”

Effects of pausing:

  • The Bot no longer accepts new tasks
  • Messages received through Gateway channels are not processed
  • Scheduled tasks do not fire
  • All existing configuration, memories, and history are preserved

Resuming a Bot

  1. On the Bot list page, click the Resume button on the paused Bot’s card
  2. The Bot status returns to “Running”

After resuming, the Bot can immediately receive new tasks and messages.

Deleting a Bot

Deleting a Bot is an irreversible operation. Please proceed with caution.

Steps

  1. Go to Bot detail page -> Settings Tab
  2. Scroll to the Danger Zone at the bottom of the page
  3. Click Delete Bot
  4. Confirm the deletion in the confirmation dialog

Impact of Deletion

  • All Bot configuration is permanently deleted
  • Resource bindings are removed
  • Memories are cleared
  • Scheduled tasks and channel configurations are deleted
  • Execution history records are preserved (for audit purposes)

Common Use Cases

Pure Conversation Bot

The simplest use case: create a Bot without binding any resources, using only the LLM for conversations.

Configuration Tips:

  • Write a clear system prompt that defines the Bot’s role and knowledge domain
  • Choose a model suitable for conversation scenarios
  • Enable auto memory to maintain context coherence across sessions

Applicable Scenarios: Customer support assistant, knowledge Q&A, copywriting.

Operations Bot

Bind SSH and web search resources to give the Bot server operation and information retrieval capabilities.

Configuration Tips:

  • Bind SSH resources with strict permission constraints:
    • Host allowlist — Only allow access to specific servers (e.g., 10.0.1.*)
    • Command allowlist — Only allow safe commands (e.g., ls *, cat /var/log/*, systemctl status *)
    • Avoid allowing dangerous commands like rm, dd, mkfs
  • Bind Web Access resources for documentation queries and troubleshooting
  • Set a monthly budget to prevent abnormal consumption
  • Use planning or adaptive reasoning mode to ensure complex operations are planned before execution
  • Configure scheduled tasks for daily inspections

Applicable Scenarios: Server monitoring, log analysis, fault diagnosis, scheduled inspections.

Orchestrator Bot

Bind multiple other Bots as resources to automatically delegate tasks to the appropriate Bot based on user needs.

Configuration Tips:

  • Create Bot-type Resources for each child Bot
  • Bind these Resources to the orchestrator Bot and configure delegate scope constraints
  • Write a system prompt explaining each child Bot’s capabilities and applicable scenarios
  • Use adaptive reasoning mode so the orchestrator Bot can adjust strategy based on child Bot execution results
  • Note permission propagation: the orchestrator Bot’s Delegate Scope for child Bots limits the child Bots’ effective permissions

Applicable Scenarios: Cross-system task orchestration, multi-agent collaboration, complex workflow distribution.

MCP Tool Integration Bot

Connect external tool servers via the MCP protocol to gain rich tool capabilities.

Configuration Tips:

  • Create an MCP-type resource in Resource Management with the tool server’s HTTP address
  • If authentication is needed, configure an API Key or OAuth
  • The platform automatically discovers available tools on the MCP server
  • Bind the MCP resource in the Bot’s resource configuration and check the tools allowed for use
  • Through tool permission controls, you can expose only selected tools to the Bot

Applicable Scenarios: Integrating third-party SaaS platform tools, using community MCP tool servers.

Agent Bridge Bot

Bridge local tools to the platform through Monstrum Agent.

Configuration Tips:

  • Create an Agent on the Agent page and obtain the Agent Key
  • Install monstrum-agent locally and connect using the Agent Key
  • The Agent registers local MCP servers, plugins, and other tools to the platform via WebSocket
  • Registered tools appear as Resources in Resource Management
  • Bind the Agent-registered resources to the Bot and configure tool permissions

Applicable Scenarios: Accessing enterprise intranet tools, using locally-run MCP servers, bridging custom scripts.

FAQ

Bot Cannot Use Tools

Possible Causes:

  1. No resources bound — Go to Bot detail page -> Resource Config Tab and check if the corresponding Resource is bound
  2. Insufficient permissions — The allowed operations selected during binding may not include the target tool. Edit permissions and confirm the required operations are checked
  3. Expired credentials — Check if the bound Credential is still valid
  4. Resource offline — For Agent-bridged resources, check if the Agent is online; for MCP resources, check if the server is reachable

Bot Replies Are Irrelevant or Off-Target

Possible Causes:

  1. Unclear system prompt — Go to the Settings Tab and check if the system prompt clearly defines the Bot’s role and behavioral boundaries
  2. Inappropriate model selection — Complex tasks require more capable models; simple conversations don’t need the most powerful model
  3. Temperature too high — Lower the Temperature value for more stable output
  4. Missing memory context — If the Bot needs to remember previous interactions, confirm that auto memory extraction is enabled

Bot Does Not Reply

Possible Causes:

  1. No LLM provider configured — Go to the LLM Providers page and confirm at least one provider exists with a successful connection test
  2. Monthly budget exhausted — Check the cost panel in the Overview Tab; if the budget is used up, increase it or wait for the next month’s reset
  3. Bot is paused — Check if the Bot status is “Paused”; if so, click “Resume”
  4. Provider API Key expired — Click “Test Connection” on the LLM Providers page to verify

Tool Call Rejected by Permissions

Possible Causes:

  1. Operation not authorized — Edit permissions in the resource configuration and confirm the target operation is allowed
  2. Parameters out of range — Guardian validates whether tool call parameters are within the allowed range. For example, SSH command allowlist, Web3 address allowlist
  3. Role permission restriction — If bound via a role, check whether the role definition’s permission scope includes the target operation

How to See What Tools the Bot Called

Go to Bot detail page -> Execution History Tab -> click “Details” on a task. In the execution timeline, you can see complete information about each tool call, including tool name, parameters, results, and permission check status.

How to Reset Bot State

  • Clear conversation context — Click “New Conversation” in the chat interface on the Overview Tab
  • Clear all memories — Go to the Memory Tab and click “Clear”
  • Reset prompts — Go to the Prompts Tab and click “Reset to Workspace Default”
  • Full rebuild — Delete the Bot and recreate it (execution history records are preserved for audit)

Scheduled Tasks Not Firing

Possible Causes:

  1. Scheduled task is disabled — Check the status column in the scheduled task list
  2. Timezone error — Confirm the scheduled task’s configured timezone matches your expectations
  3. Bot is paused — Paused Bots do not execute scheduled tasks
  4. Budget exhausted — After the monthly budget is used up, all executions (including scheduled tasks) are suspended
Edit this page on GitHub