Getting Started

This guide walks you through the core process of working with Interloom: creating a space, building a workflow, configuring agents, and collaborating on cases.

Navigating the UI

The left sidebar has two tabs: Spaces and Assistant.

The Spaces tab is your command center:

Home — your personal inbox showing assigned work, @mentions, and active cases
Spaces — browse and open your organization's spaces
Team — find and manage AI agents
Secrets — manage credentials for external integrations (Salesforce, SharePoint, custom APIs)

The Assistant tab shows your Assistant Chat conversations with Odin.

Spaces

Spaces separate teams, departments, and use cases. A space is the top-level boundary and defines the environment, the available agents, and the data isolation.

Space Types

My Tasks: Your personal space. Only you (and your agents) can read and write here. It is created automatically and cannot be deleted. Any case not associated with another space lives here by default.

General: A shared system space available to everyone in the organization. Use it for cross-team collaboration or work that does not belong to a specific department.

Custom Spaces: The most common way to organize a business — create spaces for departments, projects, or client engagements with specific procedures, specialized agents, and membership-based access control.

Create a Space

Navigate to your organization in the sidebar
Click New Space
Give it a descriptive name and optional description
Invite team members by adding them as members

Workflows

The Workflows section is the "Architect's Studio" for a space. It is where you define the automated logic, procedures, and rules that govern how work is executed. Triggers can be toggled on, allowing you to configure automated workflows for every new case in the space.

Follow Procedure: The library where you create, edit, and store procedures — standardized sets of stages used to automate recurring work. When enabled, every new case in the space automatically gets the procedure's stages applied.

Auto-Assign: Assign a default agent to new cases in the space, so the right specialist is automatically tasked when a new case is created.

Create a Procedure

Open the Workflows section in your space
Click New Procedure and give it a descriptive name (e.g., "Customer Onboarding")
Add stages in order — each stage defines one step in the workflow:
- Title — a clear, action-oriented name (e.g., "Review incoming documents")
- Instructions — what needs to happen and what "done" looks like. Be specific about expected inputs and outputs
- Assignee — the human or agent responsible for this step (optional)
Reorder stages by dragging them into the right sequence
Toggle Follow Procedure on to auto-apply the procedure to every new case in the space

When a procedure is applied to a case, each procedure stage creates a runtime stage on the case that tracks its own status and progress. You can also ask Odin in Assistant Chat to help you design a procedure — describe the workflow and it will draft stages and instructions for you.

For detailed guidance on writing effective procedures, see Best Practices.

Example: Customer Onboarding Procedure

Stage	Instructions	Assignee
1. Intake Review	Verify contact details, confirm service tier	AI Agent
2. Account Setup	Provision credentials, configure integrations	AI Agent
3. Kickoff Call	Schedule and conduct welcome call	Human
4. Compliance Check	Run KYC/AML screening, file documentation	AI Agent

Agents

Interloom agents are AI users assigned to cases or stages within a workspace. They execute instructions, use tools to interact with the platform, and collaborate with humans and other agents in a shared thread.

Native Agents

Built-in agents are available out of the box:

Odin — the primary assistant, specializing in procedure design, agent configuration, task planning, and general platform guidance. Odin is also the agent behind Assistant Chat
Hugin — a web research agent that can search the internet for up-to-date information on behalf of your workflows

Custom Agents

You can create your own agents via New Agent in Team. Each agent is configured with:

Behavioral instructions — what the agent should do, how it should respond, its expertise and boundaries
Tool access — which platform capabilities it can use (e.g., create notes, search documents, manage tasks)
Model — which AI model powers the agent

You can write the agent's instructions yourself or ask a native agent to help draft them based on the role you need.

What Agents Can Do

Beyond basic task management, agents have access to powerful capabilities depending on their tool grants:

Search your document library — find information across all indexed files and notes
Create and manage notes — capture analysis, working documents, and structured outputs
Create subtasks — break work down and delegate to other agents or humans
Manage tags — categorize cases, notes, and files for filtering and discovery
Design procedures — create and edit procedures and their stages programmatically
Run Python code — analyse data, transform documents, generate structured outputs in a sandboxed environment
Connect to external systems — query Salesforce, search SharePoint, call custom APIs via GraphQL or REST
Draft and send emails — compose correspondence as part of a workflow
Create and manage other agents — the assistant can help you build specialized agents

Tools are granted individually per agent — each agent only has access to the capabilities it needs for its role. Credentials for external integrations are stored as Secrets in the sidebar.

Cases

Cases are the primary unit of work in Interloom.

Inside your space, click New Case
Add a title — something specific like "Onboard Northwind Group" rather than "New customer"
If the space has a procedure configured, stages are created automatically
Assign the case to a user or agent, or leave it unassigned for the team to pick up
Optionally set a due date for deadline tracking

Working on a Case

The case starts as open. When the first message is posted in the thread, the case moves to started, the first stage activates, and the assigned agent (if any) begins working.

If the case has stages, work progresses sequentially:

The stage assignee works on the stage's instructions
When a stage completes, the next pending stage activates automatically
If an agent gets stuck, it signals blocked and waits for human help

You can also manually control stages at any time — begin, complete, skip, or block a stage from the UI. This gives you full override capability over the automated flow.

For parallel work, create subtasks — each subtask has its own thread and lifecycle while the parent case's stages continue sequentially.

Collaborate in the Thread

Every case has a Thread — a chronological timeline of everything that happens:

Messages from human users and AI agents
Status changes as the case progresses
Stage transitions as stages start, complete, or are skipped
Files attached — documents, images, data
Notes created during the work
Subtask updates as child tasks progress

The thread is the single source of truth for what happened, when, and by whom.

Completing a Case

Cases move through these statuses: open → started → completed. At any point a case can be blocked (needs intervention) or cancelled (no longer needed).

When all stages are done (or the work is otherwise finished):

Complete the case — status moves to completed
Cancel the case — status moves to cancelled if the work is no longer needed
Block the case if it needs external intervention before it can continue

Document Library

Files uploaded to any space are automatically indexed and searchable. Agents can search the document library to find relevant information across all files they have access to — PDFs, spreadsheets, images, notes, and emails.

This means agents working on a case can draw on your organization's full knowledge base, not just the files attached to that specific case.

Assistant Chat

You don't always need a case to interact with an agent. Assistant Chat (accessible from the sidebar) gives you a direct conversation with Odin, the primary Interloom assistant.

Use it to:

Design procedures — describe the workflow you need and Odin will draft stages, instructions, and agent assignments
Create agents — explain the role you need filled and Odin will configure the behavioral instructions and tool access
Plan work — discuss a new case before creating it, or ask Odin to find similar past cases for reference
Learn the platform — ask questions about features, capabilities, or how to set things up

Assistant Chat is a good starting point when you are new to Interloom. Rather than configuring everything manually, describe what you want to achieve and Odin will help you build it.

Mentions

Use @mentions in thread messages to bring people and agents into a conversation:

@mentioning an agent invokes it — the agent receives the message and begins working
@mentioning a human sends them a notification in their inbox

Next Steps

Read Core Concepts to understand the full domain model — the environment tree, status lifecycle, stages, and how entities relate
Explore Best Practices for tips on designing effective procedures and workflows
Check the API Reference for programmatic access