The Ultimate Company Brain Setup Checklist for Teams and Agents

Approximately 80% of enterprise data is unstructured, consisting of documents, images, and emails that remain unindexed and hidden from automated tools [IBM 2026 Report]. This unstructured information is the single largest bottleneck for modern teams attempting to integrate artificial intelligence into their operations. Traditional file repositories act as passive digital storage bins. They hold files but understand nothing of their contents, which forces human employees and AI agents to spend hours manually searching for specific documents, contract terms, or project specs.

When organizations attempt to query this unstructured storage, the results are typically fragmented. Search tools based on basic keyword matching return irrelevant files, while large language models connected to raw storage systems frequently hallucinate because they lack clear context and metadata boundaries. A team cannot expect an AI agent to execute tasks autonomously if the agent cannot find the grounding facts required to make decisions. The inability to retrieve accurate documentation leads to execution errors, broken workflows, and a lack of trust in agentic systems.

The solution is setting up an active company brain. A company brain is a centralized, intelligent workspace where human team members and AI agents share access to the same document pool. In this environment, files are not just stored. They are read, indexed, and organized on arrival, converting raw document folders into a queryable data network. This setup bridges the gap between manual storage and automated execution, enabling developers to build agentic workflows that operate on grounded, verified corporate facts.

Setting up the inbound pipeline is the first stage of building a company brain. Companies must choose how to collect files from their existing systems. While some teams try to write custom API scripts or rely on local folder copying, these methods are fragile and hard to maintain. Other platforms require continuous sync tools that frequently break or cause replication conflicts. Developers must avoid these fragile sync scripts and instead establish direct ingestion routes from corporate source repositories.

Instead of using continuous sync setups, organizations should use cloud import workflows that pull files directly from existing repositories using secure OAuth connections. With cloud import, teams can connect directly to Google Drive, Microsoft OneDrive, Dropbox, or Box, importing complete folder structures without intermediate local storage operations. For web-based documents, APIs, and public guidelines, teams can use web import tools to pull resources directly from any public URL into the workspace.

Setting up an AI-ready workspace takes under ten minutes with modern tools. During this initial setup, developers should establish clear directories for different content types. A typical directory structure includes folders for product manuals, sales contracts, customer support transcripts, and engineering briefs. Upload limits vary by plan, so check your plan's storage details before bulk-importing large archives. By structuring folders logically within shared workspaces, teams establish clear boundaries that both humans and agents can navigate.

Once files are ingested into the workspace, they must be indexed for retrieval. Traditional search systems rely on simple keyword matching, which fails when users ask questions about conceptual topics. An intelligent company brain solves this by using two distinct layers for document retrieval: semantic search for unstructured questions and metadata extraction for structured queries.

The first layer is Intelligence Mode. When Intelligence Mode is enabled on a workspace, every imported document is automatically indexed. This enables hybrid search, which combines exact full-text matching with semantic retrieval to understand the meaning behind a query. When a human team member or an agent searches the workspace, the system retrieves the exact matching passages and generates a written response. Every answer includes page-level citations pointing directly to the source documents, ensuring that all claims can be verified.

The second layer is Metadata Views. While Intelligence Mode answers natural-language questions, Metadata Views turn documents into queryable tables. Users can describe the fields they want extracted in natural language, and the AI designs a typed schema. This schema supports seven distinct field types: Text, Integer, Decimal, Boolean, URL, JSON, and Date & Time. For example, a legal workspace can extract contract dates and counterparty names, while a finance workspace can extract invoice line items and totals. These extracted fields populate a sortable, filterable spreadsheet, allowing agents to query files by metadata value (such as finding contracts where the renewal date is before a specific target). This structured extraction layer is documented in detail on the document data extraction product page. By combining semantic search with structured views, teams ensure that all data is accessible.

A company brain must support both human team members and autonomous AI agents. Human users interact with the workspace using a web dashboard, while developer tools and AI agents access the repository programmatically. Setting up this bridge requires configuring secure access keys and communication protocols. Agents should not be treated as separate silos but as integrated team members that share the workspace environment.

Developers can grant agents long-lived, scoped API keys to authenticate requests. For desktop applications and local workflows, agents use PKCE-based OAuth login, allowing authentication without exposing user passwords. Once authenticated, agents connect to the workspace using the Fast.io Model Context Protocol (MCP) Server. The MCP server exposes a consolidated toolset for workspaces, storage, AI, and workflow operations, with access points via Streamable HTTP at /mcp or legacy Server-Sent Events (SSE) at /sse. Documentation for this connection is hosted on the developer storage page.

To maintain security, administrators must define granular permissions at the workspace, folder, and file level. Agents inherit the permissions of their tokens, preventing them from accessing sensitive information like payroll spreadsheets or employee records. Within the workspace, humans and agents can collaborate in real time using Collaborative Notes. These notes support multiplayer editing with visible cursors, allowing an agent to draft a document while a human co-editor refines it. Every edit is saved in the version history, and notes are indexed automatically to serve as grounding context.

The final phase of setting up a company brain is configuring access controls, event notifications, and ownership handoffs. Rather than running manual checks, teams should build event-driven awareness into the workspace. When a file is uploaded, the workspace fires a webhook notification, allowing external services or AI agents to react to the change. Developers can register webhook listeners that trigger downstream processing pipelines as soon as files are added, and the live activity feed lets human team members watch uploads, edits, and comments happen in real time without checking folders manually.

Administrators should also lock down granular permissions before opening a workspace to agents. Permissions can be set at the organization, workspace, folder, or file level, so an agent's scoped API key only reaches the folders it needs. Every operation, human or agent, is recorded in the append-only, immutable audit log, preserving a complete chain of custody for compliance and debugging.

For service providers and developers, the ownership transfer flow simplifies deployment. An agency or developer can sign up for a free user account, build the organization, configure the workspaces, and import the client files during the fourteen-day trial. Once setup is complete, the developer sends an ownership transfer claim link to the human client. The client logs in, creates or joins the organization, enters their credit card, and begins a paid subscription. The agent can retain administrative permissions to manage the backend, while the client assumes financial and operational ownership.

Build your company brain in under ten minutes

Launch an intelligent workspace with built-in semantic search, automated metadata extraction, and developer-friendly MCP access. Starts with a 14-day free trial.

Start 14-Day Trial

To deploy a company brain, teams should follow a structured checklist that spans configuration, ingestion, and agent integration. The following steps outline the setup process.

1. Create the Organization and Start the Trial. Establish a user account and set up a new organization. Every organization runs on a paid subscription, which starts with a fourteen-day free trial that requires a credit card. Choose a plan based on storage and seat requirements: Starter plan at $29 monthly, Business at $99 monthly, or Growth at $299 monthly. Details on these subscriptions are listed on the pricing page.

2. Ingest Existing Files. Connect existing storage repositories using OAuth-based Cloud Import. Ingest files from Google Drive, Dropbox, OneDrive, or Box, keeping the original folder structures. Set up folders for distinct departments or project teams.

3. Configure Intelligence Mode. Turn on Intelligence Mode for the workspaces. Ensure that all uploaded text files, PDFs, and spreadsheets are indexed for semantic search and Q&A chat.

4. Set Up Metadata Views. Define extraction columns in natural language for structured files. Build Views to extract invoice totals, counterparty names, or project dates. Link these tables to your automated scripts using the document data extraction product page guidelines.

5. Connect AI Agents. Generate scoped API keys for your AI agents. Configure your local models or LLM frameworks to query the workspace using the Fast.io MCP Server via the streamable HTTP endpoint at /mcp.

6. Set Up Permissions and Audit Logging. Define granular permissions at the organization, workspace, folder, and file level so each agent's API key only reaches the folders it needs. Confirm the append-only audit log is capturing every file operation, so every action stays traceable for compliance and debugging.