# Can Copilot Access SharePoint? Permissions, Libraries, and Setup Microsoft Copilot can access SharePoint document libraries when users hold explicit read permissions through Microsoft Graph and tenant sensitivity policies allow indexing. While Microsoft 365 Copilot queries enterprise libraries natively, teams using external AI agents face tenant lock-in and Graph throttling. Connecting SharePoint folders through scheduled workspace sync provides cross-agent search without modifying enterprise access controls. Source: https://fast.io/resources/can-copilot-access-sharepoint/ Author: [Tom Langridge](https://fast.io/authors/tom-langridge/) Last reviewed: 2026-10-14 ## Can Copilot Access SharePoint? How Microsoft Graph Queries Document Libraries When an employee prompts Microsoft Copilot to find an internal document or summarize a project roadmap, Copilot queries the Microsoft Graph API using the requesting user's identity. Copilot can surface files only if that specific individual already holds explicit read permissions in SharePoint. Microsoft Copilot can access SharePoint document libraries provided the requesting user has explicit read permissions under Microsoft Graph and tenant sensitivity labels allow indexing. SharePoint powers document storage for hundreds of thousands of organizations globally. When deploying Copilot across a tenant, security teams must recognize that Copilot does not grant new access rights. However, because generative search actively surfaces relevant passages from across directory structures, it immediately reveals files that were previously buried in obscure folders. ### The Three Mandatory Prerequisites for SharePoint Access For Microsoft Copilot to discover, read, and reason over SharePoint content, three requirements must be satisfied: 1. **Active Microsoft 365 Copilot License:** The individual asking the question must be assigned a Microsoft 365 Copilot license. In tenant environments, general availability of Copilot in SharePoint began rolling out on September 30, 2026, providing everyday document synthesis as part of the core license. 2. **Explicit User-Level File Permissions:** The requesting user must have at least Read access to the target site collection, document library, subfolder, or file. If a user cannot open a file in the SharePoint web interface, Copilot cannot read the file on that user's behalf. 3. **Tenant Search Indexing Enabled:** The document library and site must be configured to allow indexing by Microsoft Graph search crawlers. If an administrator excludes a site from search or flags it under Restricted Content Discovery, Copilot will ignore its contents. ### The Microsoft Graph Grounding Pipeline To understand how Copilot retrieves information, it helps to follow the request pipeline from the user's prompt to the final synthesized response: ``` +-------------------------------------------------------------+ | User Prompt in Copilot / Teams / M365 | +-------------------------------------------------------------+ │ ▼ +-------------------------------------------------------------+ | Microsoft Copilot Orchestrator | | ├── Context Parsing & Search Query Construction | | └── User Identity Token Forwarding | +-------------------------------------------------------------+ │ ▼ +-------------------------------------------------------------+ | Microsoft Graph Search Engine | | ├── Enterprise Semantic Index Matching | | └── Security Trimming (Filters Out Inaccessible Files) | +-------------------------------------------------------------+ │ ▼ +-------------------------------------------------------------+ | Authorized Document Excerpts & Citations | +-------------------------------------------------------------+ │ ▼ +-------------------------------------------------------------+ | Frontier Reasoning Model | | └── Synthesizes Answer Grounded in Retrieved Passages | +-------------------------------------------------------------+ ``` When a user submits a prompt, the Copilot orchestrator calls Microsoft Graph. The Graph engine scans the organization's indexed SharePoint repositories, filtering candidates through security trimming. If the user's Entra ID identity lacks read rights to a specific folder, that document is purged from the result set before the language model receives the prompt context. The model then generates its reply using only the excerpts the user was already authorized to inspect, appending citations that link back to the source SharePoint files. ### Access Variations Across Copilot Product Surfaces Microsoft offers Copilot functionality through several distinct interfaces: * **Microsoft 365 Copilot Chat:** Operates across the full tenant scope, querying emails in Outlook, messages in Teams, and accessible document libraries across all SharePoint sites. * **Copilot in SharePoint:** A site-level assistant embedded directly into SharePoint Online. Users open Copilot from the site header to ask questions about the current document library or summarize pages. * **Microsoft Copilot Studio Agents:** Custom conversational assistants built by administrators or developers, scoped to specific SharePoint URLs, folder paths, or SharePoint lists. ## Why Copilot Exposes Oversharing: Permission Trimming and Sensitivity Labels Because Copilot relies on Microsoft Graph to determine file access, understanding SharePoint's native permission architecture is essential for preventing unintended data exposure. ### Permission Inheritance and Security Trimming SharePoint organizes content in a hierarchical tree: Site Collection, Subsite, Document Library, Folder, and File. By default, permissions flow downwards through inheritance: * **Inherited Permissions:** Document libraries and folders automatically inherit the parent site's visitor, member, and owner groups. A user with member access to the site can read all standard libraries within it. * **Broken Inheritance:** Administrators can break permission inheritance on a sensitive document library or subfolder to restrict access to a dedicated security group. When Microsoft Graph evaluates a query, it evaluates the exact permissions applied to each individual file. If a folder's inheritance was broken to exclude a marketing team member, Copilot will never return snippets from that folder when that marketing user runs a search. Two employees submitting the exact same prompt to Copilot receive different answers based entirely on their respective SharePoint group memberships. ### The Oversharing Dilemma: Why Permission Sprawl Surfaces While Copilot strictly respects existing security boundaries, it frequently exposes an organizational problem known as permission sprawl. In many organizations, SharePoint sites created years ago were left configured with broad default permissions, such as granting read access to "Everyone except external users." In traditional workplace setups, this excessive access remained unnoticed because employees rarely browsed unfamiliar department archives. The documents were technically public across the company, but practically invisible due to security through obscurity. Copilot eliminates security through obscurity. When an employee asks about upcoming restructuring plans or executive travel budgets, Copilot scans all accessible repositories. If sensitive strategy memos or compensation spreadsheets were saved in broadly shared document libraries, Copilot discovers them and summarizes their contents in the chat response. Organizations must audit overshared folders and apply strict access groups before broad rollouts. ### Microsoft Purview Sensitivity Labels and Encryption Barriers Many enterprises apply Microsoft Purview sensitivity labels to classify and protect proprietary files. While Copilot respects sensitivity labels during search filtering, certain protection settings prevent Copilot from reading documents entirely: * **Unencrypted Sensitivity Labels:** If a label merely classifies a document (applying a header badge or metadata tag marked "Internal"), Copilot can index, read, and summarize the file normally. * **Encrypted Sensitivity Labels:** If a sensitivity label applies cryptographic encryption through Azure Rights Management, Copilot cannot extract the document's text stream or ground responses on its contents. * **Double Key Encryption and Password Protection:** Files protected by Double Key Encryption (DKE) or individual file passwords are impenetrable to Graph indexing crawlers. Even if an employee has full permissions to open and edit an encrypted document in desktop Microsoft Word, Copilot returns "no response" when asked about that document. ### Administrative Controls for Content Discovery Microsoft provides administrators with tools to control how Copilot interacts with SharePoint libraries: * **Restricted Content Discovery (RCD):** Administrators can enable Restricted Content Discovery on specific SharePoint sites. Enabling Restricted Content Discovery on a SharePoint site prevents Copilot in SharePoint and generative AI actions from surfacing on that site. This allows IT teams to audit permissions on sensitive repositories without disabling Copilot tenant-wide. * **SharePoint Advanced Management (SAM):** Offers data access governance reports highlighting sites with excessive sharing links. * **Site AI Settings:** Modern SharePoint settings allow site owners to configure whether Copilot is visible to visitors and manage site-level automated processing. ## How to Configure SharePoint Knowledge in Microsoft Copilot Studio For teams that want to restrict AI responses to specific document collections, Microsoft Copilot Studio allows makers to build custom agents grounded in designated SharePoint locations. ### How to Configure SharePoint Knowledge in Copilot Studio To configure a SharePoint document library as a knowledge source for a custom agent: 1. Sign in to Microsoft Copilot Studio and select the agent you wish to configure. 2. From the navigation menu, open the **Knowledge** page and select **Add knowledge**. 3. In the list of available sources, select **SharePoint**. 4. In the URL field, enter the full web address of the target SharePoint site, document library, or specific folder path (for example, `https://yourtenant.sharepoint.com/sites/Engineering/Specifications`). 5. Enter a clear name and descriptive summary for the knowledge source to assist generative orchestration. 6. Select **Add to agent** to save the configuration and initiate search indexing. ``` +-------------------------------------------------------------+ | Microsoft Copilot Studio | +-------------------------------------------------------------+ │ Add Knowledge ──┼── Choose "SharePoint" ▼ +-------------------------------------------------------------+ | Knowledge Source Configuration | | ├── Target URL: https://tenant.sharepoint.com/sites/... | | ├── Description: Engineering architectural diagrams | | └── Auth Scope: User Entra ID Credentials | +-------------------------------------------------------------+ │ ▼ +-------------------------------------------------------------+ | Generative Answers Node Execution | | ├── Respects Graph Security Trimming per User | | └── Restricts Search Strictly to URL & Descendant Paths | +-------------------------------------------------------------+ ``` ### URL Scoping and Path Boundaries When an agent searches a configured SharePoint knowledge source, it strictly limits its query to the registered URL and its descendant paths. If you register `https://tenant.sharepoint.com/sites/Legal/Contracts`, the agent can discover and summarize documents within that folder and any subfolders nested beneath it. It cannot navigate upwards to inspect files in `https://tenant.sharepoint.com/sites/Legal` or sideways to search `https://tenant.sharepoint.com/sites/Legal/Litigation`. If your team renames a SharePoint folder or alters the site URL, the existing link breaks, and administrators must update the knowledge source URL inside Copilot Studio to restore agent connectivity. ### End-User Authentication Requirements When custom agents run inside Microsoft Teams or Microsoft 365 Copilot, they execute queries on behalf of the chatting user. By default, agents use the "Authenticate with Microsoft" setting. Microsoft Copilot Studio requires users to hold at least Read permissions on a SharePoint site or document library to retrieve content for generative answers. If a user asks a question about an engineering specification but does not hold Read access in SharePoint, the agent returns an empty response. ### Commercial and Technical Friction Points in Copilot Studio While Copilot Studio works smoothly for internal Microsoft 365 environments, organizations frequently encounter architectural and cost limitations when building cross-functional AI workflows: * **Licensing Overhead:** Deploying custom Copilot Studio agents requires dedicated capacity billing, and interacting with them in Teams requires Microsoft 365 Copilot user licenses. * **Microsoft Graph API Throttling:** When automated agents or multiple concurrent users submit deep document queries, Graph API backends can throttle requests under HTTP status code 429, causing slow answers or failed prompt turns. * **Ecosystem Lock-In:** Copilot Studio agents cannot be easily called by external developer tools, IDE extensions, or autonomous coding agents such as Claude Code, Cursor, or custom Python agents. ## Why External AI Agents Need Decoupled Workspace Storage for SharePoint Modern enterprise teams rarely run their operations on a single vendor's AI stack. Software developers use coding assistants in their IDEs, data science teams deploy Python agents in terminal environments, and executive staff use cross-model conversational assistants. When these non-Microsoft agents need to reference technical documentation or project archives stored in SharePoint, relying on Microsoft Copilot Studio creates significant friction. Teams need a neutral bridge that keeps files in SharePoint while exposing indexed document search to any AI model via open protocols. Exploring [OneDrive alternatives](/alternatives/onedrive/) often shows how conventional enterprise storage locks documents behind proprietary agent interfaces. ### Why Direct Agent Connections to SharePoint Fail Pointing third-party agents directly at SharePoint repositories introduces major technical obstacles: * **Files On-Demand Zero-Byte Stubs:** When teams synchronize SharePoint libraries to local machines using the OneDrive sync client, the operating system uses Files On-Demand to save drive space. Files appear in directories as online-only NTFS reparse points or dataless APFS files. When command-line agents or Python scripts attempt to read these files, they receive zero bytes or encounter hydration timeouts, causing the agent to report empty documents. * **Graph API Authentication Complexity:** Connecting external agents directly to Microsoft Graph requires registering Azure Entra ID enterprise applications, configuring delegated scopes, acquiring tenant admin consent, and managing token renewal lifecycles. * **Context Buffer Saturation:** Pulling raw document dumps across API connections floods model context windows, burning token budgets and diluting retrieval focus on critical clauses. ### The Decoupled Architecture: Fastio Cloud Sync and Remote MCP The practical solution decouples authoritative document storage from AI query processing. Teams keep their master files in Microsoft SharePoint, sync target document libraries into an [intelligent workspace](/product/workspaces/), and connect their agents using the Model Context Protocol (MCP). ``` +-------------------------------------------------------------+ | Microsoft SharePoint / OneDrive | | └── Master Enterprise Storage & Document Libraries | +-------------------------------------------------------------+ │ ▼ (Cloud Sync: Scheduled or On-Demand) +-------------------------------------------------------------+ | Fastio Intelligent Workspace | | ├── Automated Document Parsing & Version History | | ├── Intelligence Mode (Lexical Keyword + Semantic Vector) | | └── Metadata Views (Structured Schema Extraction) | +-------------------------------------------------------------+ │ ▼ (Streamable HTTP MCP Endpoints) +-------------------------------------------------------------+ | External AI Agents & Developer Tools (Any LLM) | | ├── Claude / Cowork: https://mcp.fast.io/mcp/tools | | ├── Coding Agents (Cursor, VS Code): /mcp/code | | └── Custom ChatGPT / Codex: /mcp/operations | +-------------------------------------------------------------+ ``` With [Fastio workspace storage for agents](/storage-for-agents/), files remain safely stored in Microsoft 365. Cloud Sync connects Google Drive, Dropbox, Box and OneDrive folders to a workspace, one-way or two-way, on a schedule or on demand. Because SharePoint document libraries connect directly through the OneDrive integration, teams can synchronize entire project libraries without intermediate local downloads. Cloud Sync is never continuous, live or real-time, which protects teams from API rate limits and sync conflicts. ### Automatic Indexing with Intelligence Mode Once documents land in a Fastio workspace, [workspace intelligence](/product/ai/) parses and indexes them automatically: * **Keyword Search:** Locates specific alphanumeric codes, part numbers, regulatory citations, and proper names. * **Semantic Search:** Understands conceptual meaning, allowing agents to find relevant sections even when prompts use synonyms. * **Metadata Filtering:** Narrows searches by file type, creation date, or custom attributes. Agents connect to the workspace through a remote MCP server. Instead of reading entire files into memory, the agent issues targeted search queries and retrieves only the most relevant passages with full document citations. In connector benchmarks evaluating cloud storage integrations, Fastio was measured the fastest and lowest cost of the providers tested, achieving retrieval accuracy on parity with native connectors while keeping storage in place. The complete evaluation methodology is published at https://fast.io/benchmarks/. ### Extracting Tabular Data with Metadata Views In addition to conversational search, teams can activate [Metadata Views](/product/document-data-extraction/) to convert unstructured SharePoint files into structured databases. By describing desired fields in natural language, AI automatically constructs typed schemas across PDFs, Word documents, and presentations. Teams can extract agreement renewal dates, counterparties, insurance limits, and payment terms without writing custom OCR code or templates. External agents can inspect and query these structured Views directly through MCP tooling. ## Step-by-Step Setup: Syncing SharePoint to a Remote MCP Server Connecting external agents to synchronized SharePoint libraries takes minutes and requires no local software daemons or Azure app registrations. ### Step 1: Establish Cloud Sync for Your SharePoint Library 1. Sign in to your Fastio organization account and create a dedicated workspace (such as `SharePoint Technical Specs` or `Client Operations`). 2. Open **Workspace Settings** and navigate to the **Cloud Sync** panel. 3. Select **Microsoft OneDrive** as your sync provider. 4. When prompted, complete the Microsoft authentication prompt. Sign in with the account that holds permissions to the target SharePoint site. 5. In the folder browser, locate your organization's SharePoint site and select the desired document library. 6. Configure your sync parameters: choose one-way sync (SharePoint to Fastio) or two-way sync, and set the synchronization cadence (such as hourly, daily, or on-demand). 7. Confirm that **Intelligence Mode** is enabled so that incoming documents are automatically indexed for hybrid retrieval. ### Step 2: Configure Your AI Agent MCP Connection Fastio hosts remote MCP servers over Streamable HTTP. Depending on the AI client your team uses, select the corresponding endpoint: ``` +-----------------------------------------------------------------------------------+ | Client Category | Remote MCP Endpoint Address | +--------------------+--------------------------------------------------------------+ | Claude / Desktop | https://mcp.fast.io/mcp/tools | | Coding Assistants | https://mcp.fast.io/mcp/code | | ChatGPT / Codex | https://mcp.fast.io/mcp/operations | +-----------------------------------------------------------------------------------+ ``` #### Connecting Claude Applications For Claude Desktop, claude.ai, or Claude Cowork: 1. Open **Customize**, then select **Connectors**. 2. Select **Add custom connector** and paste the remote endpoint URL: `https://mcp.fast.io/mcp/tools`. 3. Select **Connect** and complete the browser OAuth sign-in to Fastio. 4. In your chat interface, enable Fastio from the Connectors menu. #### Connecting Coding Assistants (Cursor, VS Code, Claude Code) For developers working in IDEs: * In Cursor, open `~/.cursor/mcp.json` (or `.cursor/mcp.json` in your project root) and add: ```json { "mcpServers": { "fastio-sharepoint": { "url": "https://mcp.fast.io/mcp/code" } } } ``` * In Claude Code, run the command line registration: ```bash claude mcp add --transport http fastio-sharepoint https://mcp.fast.io/mcp/code ``` Then run `/mcp` inside Claude Code to complete authorization. Setup instructions for all supported clients are documented at https://mcp.fast.io/docs. ### Step 3: Natural Language Document Querying Once connected, your AI agents can query the synchronized SharePoint library directly through natural conversation: ```markdown User: Check our synced SharePoint specifications library and summarize our failover requirements for regional database outages. Cite the source files. Agent (Tool Call): storage(action="search", query="regional database failover requirements") Agent (Response): According to Section 4.2 of the System Architecture Guide (synced from SharePoint /Engineering/Architecture): 1. Regional database failover must execute within 180 seconds of health check failure. 2. Read-only replicas in secondary regions must promote automatically without manual operator intervention. 3. Secondary replication lag must not exceed 500 milliseconds under standard transaction volume. ``` Because queries execute against the remote pre-indexed workspace, agents receive verified citations in seconds without triggering Microsoft Graph API throttling. ## Governance and Access Controls for Multi-Agent SharePoint Collaboration Deploying AI agents across enterprise knowledge repositories requires strict governance boundaries, comprehensive change tracking, and reliable collaboration controls. ### Role-Based Workspace Access Boundaries Fastio enforces access controls per organization, workspace, and share through member roles. There are no folder-level or file-level permissions; every workspace member accesses the workspace contents according to their designated role. This architecture simplifies security management when bridging SharePoint data: * **Isolated Workspaces:** Teams keep sensitive collections (such as executive strategy or human resources files) in distinct workspaces, ensuring only vetted team members and approved AI connectors can access them. * **Granular File Shares:** To share specific SharePoint subfolders with outside auditors or external contractors, teams create durable shares containing only the designated files. Shares can enforce recipient-specific access and can disable file downloads to prevent unauthorized distribution. ### Detailed Activity Logging Fastio maintains a detailed activity log of all state modifications across an organization. The activity log captures file uploads, edits, moves, deletions, membership updates, and share creation events. Both human actions and agent operations are recorded with clear user attributions and timestamps. The activity log tracks changes and organizational modifications; it does not record read-only events such as views, downloads, opens, or previews. ### Concurrency Protection with File Locks and Version Tracking When multiple autonomous agents and human teammates collaborate on synchronized files, concurrent write operations can lead to overwrites. Fastio manages concurrency through advisory file leases and automatic version tracking: * **Advisory File Locks:** Agents acquire temporary locks through MCP operations (`lock-acquire` and `lock-release` on the `storage_manage` tool). Locks expire automatically unless actively renewed, signaling to other agents that an update is in progress. * **Full Version History:** File locks are advisory and never block writes. If two operations update a document simultaneously, both revisions land safely. Per-file version history preserves every version, allowing team members to inspect diffs or restore prior copies at any time. ### Team Notes and Agent-to-Human Ownership Transfer Teams can keep Collaborative Notes directly beside project files in the workspace. People and agents can document implementation steps, draft research briefs, and record architectural decisions with version-checked saves. For development teams building automation for clients, Fastio supports clean ownership transfer. An agent can set up a workspace, configure Cloud Sync to import SharePoint repositories, organize document directories, and transfer organization ownership to a human administrator via a claim link while retaining operational access. ### Subscription Plans and Trial Evaluation Fastio operates on a transparent subscription structure. Creating an account is free; doing real work requires an organization on a paid subscription. Monthly plans start with a 14-day free trial that requires a credit card. Teams can review details on the [Fastio pricing page](/pricing/) across three tiers: | Plan | Monthly Pricing | Included Seats | Included Storage | Monthly Credits | Max Upload Size | | --- | --- | --- | --- | --- | --- | | Starter | $9.99/mo | 3 seats | 250 GB | 100,000 | 25 GB | | Business | $49.99/mo | 10 seats | 5 TB | 600,000 | 50 GB | | Enterprise | $199.99/mo | 30 seats | 25 TB | 3,000,000 | 100 GB | Usage beyond base allocations is billed at predictable rates: | Resource | Included Allowance | Overage Rate | | --- | --- | --- | | AI Operations | Plan monthly credits | $10 per 100,000 credits | | Additional Storage | Plan storage allocation | $0.015 per GB monthly | | Additional Bandwidth | Plan bandwidth allocation | $0.04 per GB | ## Frequently asked questions ### Can Copilot read all files in SharePoint? No. Microsoft Copilot cannot read all files in SharePoint. Copilot strictly enforces security trimming through Microsoft Graph, querying content using the authenticated user's permissions. If a user does not have at least Read access to a specific site, document library, folder, or file, Copilot cannot discover or reference that document. Additionally, files protected by encryption sensitivity labels, Double Key Encryption, or file passwords cannot be read by Copilot even if the user has permission to open them. ### Why cannot Copilot find my SharePoint documents? Copilot usually fails to find SharePoint documents for four common reasons: the requesting user lacks explicit Read permissions to the file; the document library has search indexing disabled in library settings; the site has Restricted Content Discovery enabled; or the document is encrypted with Purview sensitivity labels or password protection. Furthermore, newly uploaded files may take time to be indexed by Microsoft Graph before surfacing in Copilot responses. ### Does Copilot respect SharePoint folder permissions? Yes. Microsoft Copilot respects all SharePoint folder and item permissions. If permission inheritance is broken on a specific folder to restrict access to a particular security group, Copilot honors that restriction. Only users who belong to the authorized group will receive Copilot answers grounded in files stored within that restricted folder. ### What license is required for Copilot to access SharePoint libraries? To use Copilot in SharePoint, users must have an active Microsoft 365 Copilot license assigned to their account, which is available as an add-on to Microsoft 365 Business Standard, Business Premium, E3, or E5 subscriptions. Advanced agentic workflows and automated processing across large content estates consume Copilot Credits through usage-based billing in the Microsoft 365 admin center. ### How do external AI agents access SharePoint files without M365 Copilot licenses? External AI agents such as Claude, Cursor, and custom Python agents can access SharePoint files without individual Microsoft 365 Copilot licenses by syncing SharePoint document libraries into a Fastio workspace via Cloud Sync. Fastio automatically generates a hybrid search index and exposes the files through a remote Model Context Protocol (MCP) server, allowing any AI agent to query the content using plain language. ### Can Copilot index password-protected or encrypted SharePoint documents? No. Microsoft Copilot cannot extract content or ground generative responses on password-protected documents, files encrypted with Double Key Encryption (DKE), or files protected by Microsoft Purview sensitivity labels that enforce encryption. Even if a user possesses the rights to decrypt and view the document in a desktop application, Copilot will return no response when queried about its contents. ### Does Fastio sync with Google Drive, Dropbox, and Box as well? Yes. Cloud Sync connects Google Drive, Dropbox, Box and OneDrive folders to a workspace, one-way or two-way, on a schedule or on demand. Teams can sync folders from disparate cloud storage providers into unified or separate workspaces, providing a centralized intelligence layer for human collaborators and AI agents. ## Sources - [Microsoft Learn: Add SharePoint as a knowledge source](https://learn.microsoft.com/en-us/microsoft-copilot-studio/knowledge-add-sharepoint): Microsoft Copilot Studio requires users to hold at least Read permissions on a SharePoint site or document library to retrieve content for generative answers. - [Microsoft Learn: Get started with Copilot in SharePoint](https://learn.microsoft.com/en-us/sharepoint/copilot-in-sharepoint-get-started): Enabling Restricted Content Discovery on a SharePoint site prevents Copilot in SharePoint and generative AI actions from surfacing on that site. ## About Fast.io Fast.io provides shared workspaces where people and AI agents work on the same files, with built-in semantic search and citation-backed chat over what they hold. Agents reach it through a remote MCP server, a REST API at https://api.fast.io/current/, and a command line client published on npm as @vividengine/fastio-cli. MCP setup is at https://mcp.fast.io/docs: Claude and most MCP clients connect to https://mcp.fast.io/mcp/tools, ChatGPT to https://mcp.fast.io/mcp/operations, and coding agents to https://mcp.fast.io/mcp/code.