Docs

What is Anchor?

Anchor is cloud infrastructure for AI agents. You connect an MCP-compatible agent — Claude, ChatGPT, Gemini, Grok, Codex, or any other — and it creates, hosts, and manages apps, tables, and files on your behalf. You talk to your agent; Anchor handles the rest.

Think of it as Google Drive, but the items are a new generation: fully functional apps, live interactive tools, structured tables, and any file type you need. Everything lives at anchor.cc and works in any browser on desktop or mobile.

Item types

Every item in Anchor is a file. There are three broad kinds — apps, tables, and other files — and each one has a stable URL you can link to, bookmark, or share.

Apps & online tools

Apps are fully custom, mobile-friendly web applications. Each app has two URLs:

• The running app lives at <slug>.anchor.cc — a subdomain per app. That's the link you send to users who should open and use it.
• The management page at anchor.cc/app/<appID> is where you see the app's version history, edit it with your agent, manage sharing, and review its connections to tables and files.

Anchor supports two kinds of apps:

• Data-backed apps — Backed by Anchor tables for persistent storage. Custom CRMs, ERPs, task trackers, forms, dashboards, HR tools, finance trackers, event planners — basically any software, described in plain language. The app reads and writes directly against your tables, so everything stays in sync.
• Standalone apps — Visual and informational content with no backend. Styled documentation, charts, flow diagrams, interactive presentations, slides.

Apps are versioned. Your agent proposes changes, you review, and you commit — the previous version stays in history so you can roll back.

Tables

Tables are structured, Excel-like data at anchor.cc/table/<tableID>. Your AI agent is fully connected to them, which means you can:

• Create and edit schemas — add or rename columns, change types, reorganize.
• Query, filter, and sort rows in natural language. No formulas, no SQL.
• Transform and analyze — group, aggregate, pivot, join, or derive new columns from existing ones.
• Visualize — ask for charts, summaries, and breakdowns, either inline in the agent or as a dashboard app.

Tables are also the persistence layer for data-backed apps. A single table can power multiple apps at once, so your CRM table, your pipeline dashboard, and your weekly report app all read from the same source of truth.

Files (other types)

Anything that isn't an app or a table is stored as a file at anchor.cc/file/<fileID>. Your agent uploads files for you; everyone you share them with can view them in the browser, and any file can be downloaded at anchor.cc/file/<fileID>/download. This gives you a single, unified storage layer — your personal or organizational knowledge base, fully readable by your AI agents.

Supported file types

Anchor recognizes these file types and renders them natively in the browser. Unrecognized types can still be uploaded and downloaded — they just show as generic binary files.

Uploading

Files can be uploaded by your AI agent or directly through the Anchor UI. The upload destination is always a folder — jump to any folder's upload dialog with anchor.cc/folder/<folderID>/upload.

Connections

Apps depend on tables and files to function. These dependencies are called connections. When your agent creates or edits an app, it declares which tables and files the app needs, and at what access level.

Access levels

Connections come in two flavors — the same two roles used for human access:

Consumer

Read-only. The app can query and display the data but cannot change it.

Contributor

Read and write. The app can create, update, and delete rows or content in the target file.

Lifecycle & approval

Whether a connection activates immediately depends on your permissions on the target:

• You're a contributor or owner of the table or file → the connection is granted instantly. You already have authority to let an app use it.
• You're only a consumer → the connection is requested. The target's owner or a contributor must approve before the app can use it, and can deny or revoke access at any time.

Managing connections

Every app's management page lists its outgoing connections — the tables and files it reads from and writes to — so you always know what an app touches. Every table and file shows its incoming connections — the apps that depend on it. From either side, owners and contributors can approve, deny, or revoke access.

An app is considered fully connected when all its declared dependencies are granted. Only fully connected apps serve live traffic; if a table or column is missing or pending approval, the app falls back to an inactive state until the connections resolve.

Structure: orgs, folders, files

Anchor organizes everything into three layers.

Organizations

An organization is the top-level boundary — your team, company, or any group that works together. Every org has three roles:

Owner

Full control. Can manage members, rename, delete, and transfer ownership of the org.

Manager

Can add and remove members, invite new people, and rename the org.

Member

Can see the org roster. Cannot manage members or change org settings.

Org roles are purely administrative — they control who can manage the org itself, not who can see its content. Being an owner, manager, or member of an org does not grant access to any folders or files inside it. Content access must be granted separately at the folder or file level.

Folders

All content lives inside folders. Folders are always scoped to an organization and can be nested to any depth, just like a traditional file system. Every folder has its own permissions: an owner, contributors (can view and edit), and consumers (view only).

Files

Apps, tables, and other file types all live inside folders. Each file has its own permissions too — independent of the folder it belongs to. You can share one file from a private folder, or make one file view-only inside a fully editable folder.

Sharing & permissions

Anchor has a granular, explicit sharing model. Access is never inferred — it is always granted. Share individual files, entire folders, or grant org membership. Each level has its own permissions.

Three roles, everywhere

Every folder and every file uses the same three roles:

Owner

Full control. Can edit, share, transfer ownership, and delete.

Contributor

Can view, edit, create sub-items, move, copy, and invite others.

Consumer

Can view and copy. Cannot edit, delete, or see who else has access.

Additive inheritance

Folder permissions flow down to subfolders and files. If someone is a contributor on a parent folder, they're a contributor on everything inside. The highest access level always wins — a direct consumer on a file can still be a contributor via an ancestor folder. Org membership does not propagate down; it's purely administrative.

Who can see the permissions?

Consumers see only the owner of a file and their own role. Contributors and owners see the full list of who has access — direct contributors, consumers, invitations, and any app connections.

Shared with me

Files and folders shared with you from outside your own organizations appear in a single Shared with me view. Anchor automatically consolidates upward so you see the most useful starting point, not every individual item.

Invite by email

You can invite anyone as a contributor or consumer by email, even if they don't have an Anchor account yet. They'll gain access when they sign up. Invites are sent as accept-links — anchor.cc/file/accept-invite/<inviteID> for files, anchor.cc/folder/accept-invite/<inviteID> for folders, and anchor.cc/org/accept-invite/<inviteID> for org membership.

Plans & billing

Billing in Anchor is tied to the organization, not the individual. There's no per-seat pricing: an org pays one price that covers everyone in it.

Plans are per org

Every organization is on a plan — Free, Pro, Business, or Enterprise. Upgrading or downgrading applies to that org only. If you belong to multiple orgs, each one has its own plan, its own invoice, and its own settings. See pricing for the current tiers and limits.

Limits are per org

Usage limits are enforced at the organization level and count every item created inside it. Depending on your plan, limits include:

• The number of apps and tables the org can host.
• The number of rows across all tables.
• Operations per month — each read, write, or agent action against Anchor.
• The number of team members in the org.

If you approach a limit, Anchor surfaces a warning in the org's settings page. Paid plans can top off capacity mid-month without switching tiers.

Budgets & visibility

Org owners and managers see a live usage dashboard in org settings — current consumption, rate of growth, and remaining headroom against each limit. Regular members don't see billing details but can see when an org is close to a limit so they know before an operation fails.

Shared content

Items that another org has shared with you don't count toward your org's limits; they stay billed to the org that owns them. When you collaborate across orgs, each side pays only for what it created.

Connecting your AI agent

Anchor is mostly used through your AI agent, not the Anchor UI directly. Your agent connects to Anchor over the Model Context Protocol (MCP) — the open standard for AI tool use — and gains the ability to create apps, manage tables, organize folders, upload files, and control sharing on your behalf. Any MCP-compatible agent works.

The Anchor UI at anchor.cc is for viewing, navigating, and managing what your agent has already built. Most day-to-day work happens inside your agent: you describe what you want, your agent does it in Anchor, and you review the result.

For setup instructions, the full list of supported agents, and the MCP tool reference, see the MCP guide.