How to Set Up a Company Brain in Ten Minutes: Quickstart Guide

The Hidden Complexity of Custom Retrieval Stacks

Engineering teams often assume that giving an AI access to internal knowledge requires building a custom retrieval pipeline from scratch. A typical DIY implementation requires spinning up an external vector database, provisioning an embedding service, writing custom chunking scripts for PDFs and spreadsheets, and deploying an orchestration server running LangChain or LlamaIndex.

This architecture introduces substantial technical debt. Document parsers frequently fail on complex tables or image-heavy presentations. When a policy changes or a contract is amended, updating the external vector index requires custom synchronization jobs. If the synchronization job drops a webhook or fails silently, the vector database serves stale records while the source repository holds the corrected file. Maintaining this custom infrastructure pulls engineers away from core product work and drives up recurring cloud API bills. Storage-level indexing replaces this multi-tier architecture by embedding retrieval directly into the file repository where work happens.

What Documents Belong in an Operational Knowledge Base

A company brain is not an unfiltered archive of every temporary draft or chat transcript created across an organization. To keep answers precise and prevent context clutter, organizations focus on core operational records that reflect business ground truth.

Effective company brains organize five main categories of material:

• Technical Specifications and Architecture Decisions: System designs, API schemas, deployment runbooks, and Architecture Decision Records (ADRs) that explain engineering choices.
• Standard Operating Procedures (SOPs): Step-by-step operational checklists, release schedules, customer support playbooks, and incident remediation steps.
• Customer Agreements and Vendor Contracts: Executed master service agreements, statements of work, vendor licensing limits, and procurement documentation.
• Organizational Policies and Onboarding Guides: Employee handbooks, benefits policies, security protocols, and team-specific onboarding curricula.
• Autonomous Agent Outputs: Research syntheses, competitive intelligence briefs, and structured document extracts generated by internal agents for team reuse.

Housing these documents in an intelligent workspace makes them queryable in plain English without requiring manual tagging or folder reorganization.

Workspace Creation and Automated Storage Indexing

The first step establishes a clean workspace boundary. An administrator creates an organization on Fast.io and establishes a dedicated workspace named for the company brain, such as core-knowledge or company-brain.

Within the workspace settings, turning on Intelligence Mode starts background indexing. When a PDF, Word document, markdown file or text document lands in a workspace with Intelligence on, it is queued for text extraction and indexing automatically.

Workspaces handle substantial document volumes without manual intervention. Upload limits scale with plan tiers, allowing large collections of vendor manuals, scanned contracts, and historical documentation to upload directly without splitting archives into separate chunks.

Connecting Storage Drives with Scheduled Cloud Sync

The next step brings historical records into the newly created workspace. Moving away from scattered folders does not require downloading gigabytes of PDFs to a local machine before re-uploading them.

Cloud Sync connects Google Drive, Dropbox, Box, and OneDrive folders directly to the workspace. For organizations running on Microsoft 365, SharePoint document libraries connect through the OneDrive connector. Administrators select whether the sync operates one-way to mirror reference files or two-way to keep folders aligned across environments. Sync runs on a schedule or on demand, rather than continuous or real-time polling, keeping system load predictable.

A key security principle governs this ingestion: synced files adopt the Fast.io workspace permissions, not the source drive's sharing settings. If a Google Drive folder had loose link-sharing settings, those settings do not carry over: in Fast.io, the synced copies follow the workspace's permissions. The original links in Google Drive are unchanged. The first sync of a large archive, and the indexing that follows, continue in the background after the connection is set up.

Configuring the Remote MCP Endpoint

The third step connects external AI agents to the repository. Rather than writing custom API wrappers for each LLM provider, Fast.io exposes its capabilities through the Model Context Protocol (MCP), an open standard for connecting AI clients to external context.

The Fast.io MCP server is fully hosted and remote, operating over Streamable HTTP without requiring local package installations or node runtime environments. Connecting an AI tool is a short step:

• Claude (Web, Desktop, and Mobile): Open Customize, navigate to Connectors, select Add custom connector, and paste https://mcp.fast.io/mcp/tools. Select Connect and complete the browser OAuth sign-in. In any chat conversation, activate Fastio from the plus menu under Connectors.
• Claude Code: Run claude mcp add --transport http fast-io https://mcp.fast.io/mcp/code in your terminal, then execute /mcp to authenticate.
• Cursor: Add an entry to ~/.cursor/mcp.json or .cursor/mcp.json with the server URL https://mcp.fast.io/mcp/code.
• ChatGPT and Codex: Install the Fastio plugin directly from the plugin directory, or configure a custom MCP connector pointing to https://mcp.fast.io/mcp/operations.
• Headless Agents and Scripts: When running autonomous daemons, pass a scoped Fast.io API key in the connection header as Authorization: Bearer <api key>.

Detailed setup instructions are available in the remote MCP documentation, with full tooling descriptions documented in the MCP skill guide and agent onboarding documentation.

Once the client is connected, write an Ask the Brain skill for the agent: an instruction to query the company brain workspace before drafting, planning or changing anything, and to cite the file behind every factual claim. This last piece of wiring is what makes the agent check the workspace instead of guessing.

Executing the Initial Verification Query

The final step checks that files are indexed and agents can retrieve ground truth. Testing the setup requires asking a question that depends on proprietary facts contained within an ingested document.

For example, asking 'What are our payment terms for enterprise clients under the standard master services agreement?' prompts Intelligence Mode to scan the indexed files. The retrieval engine combines full-text keyword matching with vector-based semantic search to identify relevant paragraphs across hundreds of pages.

The response returns the substantive answer along with explicit citations. Each citation links to the source document, the specific page number, and the excerpted snippet that provided the evidence. Citations let teammates check any claim against the file it came from. Teammates can click any citation to inspect the underlying file in the workspace viewer.

Workspace-Level Access Control for Agents and Humans

To maintain document security, organizations structure workspaces by domain and sensitivity. For instance, general SOPs, technical specifications, and company policies live in a company-wide workspace where all employees and general AI assistants hold Member roles. Financial ledgers, executive compensation sheets, and board minutes live in an executive workspace restricted to leadership.

When creating a new workspace, the join setting defaults to allowing any member of the parent organization to join. To keep confidential data private, administrators restrict this setting so only named individuals and designated agent accounts can enter.

Organization roles include Owner, Admin, and Member, while workspace roles add Guest and View. API keys used by autonomous agents act with the permissions of the user account that created them, narrowed by the scopes set on the key. If an agent only needs to read, the person connecting it chooses Read Only on the sign-in screen, and data-altering tools are then not offered. That removes the risk of unintended file edits or deletions through that connection.

Tracking Document Lifecycle Changes in the Activity Log

Auditing changes is critical when multiple people and automated agents share a knowledge repository. Fast.io includes a detailed activity log that records operational modifications across the organization.

The activity log tracks concrete changes: file uploads, content edits, file moves, deletions, membership updates, permission adjustments, and share creations. Both human actions and agent operations are attributed with timestamps and user identities. The log does not track read-only actions such as views, file opens, or downloads.

In addition to activity logging, Fast.io maintains full per-file version history. When an agent updates a technical document or an employee uploads an amended handbook, the previous version is preserved. If an erroneous edit occurs or an agent overwrites useful text, administrators can review the version list and restore earlier versions with a single click. This versioning framework provides a safety net that protects the integrity of the company brain.

Automating Schema Design and Document Extraction

Defining the fields to extract is quick: describe the information you want in plain language. An administrator or team lead creates a view within a workspace and enters a prompt describing the desired information. For example, a legal operations lead might specify: 'Extract counterparty name, effective date, renewal notice deadline, contract value, and governing law.'

The system generates a typed schema supporting seven field formats: Text, Integer, Decimal, Boolean, URL, JSON, and Date & Time. As files match the view criteria, the system extracts values and populates the view columns automatically.

Because extraction operates incrementally, teams can add new columns at any point without re-uploading or reprocessing the entire file archive. AI agents connecting through the Fast.io MCP server can inspect schemas, query extracted values, and use structured document metadata to trigger subsequent actions, bridging the gap between raw document storage and structured data analysis.

Maintaining Living Context with Collaborative Notes

Not all institutional knowledge arrives in polished PDF reports or formal handbooks. Much of a team's operational context develops during project execution, technical debriefs, and ongoing agent tasks.

Fast.io provides Collaborative Notes kept directly beside files within workspaces. These notes allow team members and autonomous agents to document operational insights, meeting takeaways, and workflow guidelines in clean markdown.

Collaborative Notes support live co-editing, allowing people and agents with the same note open to see each other's labeled cursors and edits as they type. An engineering agent can read an architecture document, execute a benchmarking script, and append its findings to a Collaborative Note beside the design specification. Human engineers reviewing the workspace see the agent's annotated notes alongside the source files, maintaining transparent collaboration across human and machine workflows.

Developer and Terminal Agent Integration

For technical teams, the primary value of a company brain is programmatic accessibility. Developers do not want to switch context to a web browser to search for internal documentation while writing code.

By adding Fast.io to their IDE or terminal client configuration, engineers allow their coding agents to query storage for agents directly. When a developer asks their agent to implement a new database migration or integrate an authentication provider, the agent calls the Fast.io MCP search tool to find relevant internal standards.

Because the MCP server handles authentication and retrieval through Streamable HTTP, developers work with standard tools without running local synchronization daemons. The workspace serves as persistent memory across development sessions, ensuring that context remains stable even when local development environments are reset.

Subscription Plans and Scalable Deployment

A company brain on Fast.io scales with organizational needs. Every organization runs on a paid subscription, and monthly plans start with a 14-day free trial that requires a credit card.

The table below outlines the core subscription plans and resource allocations:

Seats count organization members only, while external collaborators accessing shares do not consume account seats. Extra storage costs $0.015 per GB a month and extra bandwidth $0.04 per GB, neither using AI credits; credits cover chat, file ingestion, agent runs and document processing, with overage at $10 per 100,000 credits. More details are available on the Fast.io pricing page.