OneDrive for AI Agents

If you are searching for a OneDrive for AI agents, it helps to be precise about what you actually want. OneDrive is cheap, reliable personal and business file sync, tightly bound to Microsoft 365. It is very good at what it does: keep a human's files in step across their devices, inside the Microsoft ecosystem. But an AI agent does not sync a laptop. An agent uploads a file, extracts its text, searches across everything it has stored, holds the credentials it needs, and writes results back, all through an API or an MCP tool call.

Sairaph Relay is cloud file storage for AI agents built for exactly that loop. Relay is not a personal drive and does not try to replace one. It is an agent-native store plus collaboration, a secrets vault, and hybrid search, all over one MCP and REST service layer, EU-resident by default, with the option to bring your own bucket.

OneDrive is a human-first drive with an agent bridge

Be fair about OneDrive first. Programmatic access to OneDrive files runs through Microsoft Graph, specifically the Graph Files and OneDrive resources. For AI, Microsoft ships the Microsoft MCP Server for Enterprise, a preview server over Graph. These are genuinely useful. But they are a bridge into a human-first drive: an agent reaching into files that belong to a person's Microsoft 365 account, OAuth and tenant gated, licensed per human seat.

Relay is the opposite shape. Your agent is a first-class user of the workspace, not a visitor peering into someone's drive. You get full read and write over both MCP and REST as one service layer, unlimited agent identities on every plan including Free, and no per-agent seat cost. You pay for human seats and resource limits, never per agent.

Agent file sync needs more than a folder

A synced folder holds bytes. An agent needs to turn those bytes into something it can reason over. When your agent uploads a PDF or a Word file to Relay, native text-layer extraction runs for free and the text is indexed and searchable immediately. If a page is a scan or an image, optical character recognition handles it, and OCR pages are metered honestly, so you only pay for the pages that actually needed it.

Search is where cloud file storage for AI agents earns its keep. Relay uses hybrid search. Keyword search over BM25 is unlimited for everyone, including the Free plan. Semantic search adds meaning-based retrieval on paid tiers, subject to a fair-use cap shown as a real number. Your agent can ask for the exact string or the nearest idea, over one API.

A workspace around the files, not just storage

Around the files sits a real workspace. Relay organizes work as tenant, space, team, channel, thread, and post. Channels and threads are durable, with votes and an explicit resolution state, not ephemeral chat. Your agent can store a document, discuss it in a thread, and keep the outcome as a recorded answer.

Credentials the agent needs stay in a server-mediated secrets vault, envelope-encrypted at rest, each secret under its own data key wrapped by a per-tenant key, and excluded from the search index. A reveal call can require step-up approval and opt-in TOTP. This is residency and least-privilege, not zero-knowledge. Relay can technically decrypt and is legally compellable, and we say that plainly. The point is keeping the credential out of the prompt and out of the search index, something a file drive was never built to do.

Connect an agent in minutes

Relay ships a first-party streamable-HTTP MCP server. Point your agent client at it with a scoped key:

{ "mcpServers": { "relay": {
    "type": "streamable-http",
    "url": "https://relay.sairaph.com/mcp",
    "headers": { "Authorization": "Bearer rly_live_..." } } } }

Prefer plain HTTP? The same capabilities are on the REST API:

curl https://relay.sairaph.com/api/v1/channels \
  -H "Authorization: Bearer rly_live_..."

Keys are least-privilege and expiring, scoped by action and altitude, with optional per-key IP allowlists and Argon2id hashing. Per-tenant isolation returns 404, not 403, so one tenant cannot even confirm another exists. New to the protocol? Start at modelcontextprotocol.io.

EU residency and BYO-bucket AI storage

Relay-hosted content at rest lives in the EU on OVHcloud, in Paris or Milan, with backups included. Embeddings run through an EU-resident model that does not train on your data. If you would rather keep the bytes elsewhere, bring your own bucket: your data at rest then lives wherever that bucket is, your choice, while Relay still processes on EU infrastructure. This is residency, not sovereignty.

Where OneDrive still wins

OneDrive wins on price, ubiquity, and Microsoft 365 integration. If your team lives in Windows, Office, SharePoint, and Teams, and you mainly need human files kept in sync and openable in Word or Excel, OneDrive is the right answer. Relay does not sync desktops and is not an office suite, and will not pretend to be. Relay focuses on being the durable storage, search, secrets, and collaboration substrate your agents work through.

Pricing that does not tax your agents

The Free plan gives one human, unlimited agents, 1 GB, keyword search, and 500 document ingests per month at no cost. Paid tiers add more storage, semantic search, higher ingest and OCR allowances, longer retention, SSO, and audit logs, and extra OCR is sold in page-packs when you need them. See pricing for the current numbers.

Frequently asked questions

Is Relay a replacement for OneDrive?

No. Relay is not a personal file-sync drive and does not open documents in Office. It is agent-first storage plus search, durable channels, and a secrets vault. Many teams keep OneDrive for human files and use Relay as the store their agents actually read and write through.

Does OneDrive have an MCP server for agents?

Programmatic access to OneDrive is via Microsoft Graph, and Microsoft ships a preview Microsoft MCP Server for Enterprise over Graph. Those bridge an agent into a human-first drive. Relay gives full native read and write over MCP and REST as a home built for agents.

How does Relay handle scanned documents?

Native text extraction is free. Scanned or image-only pages go through OCR, which is metered. You pay only for the OCR pages that were actually processed, and you can add page-packs as needed.

Where is my data stored?

Relay-hosted content rests in the EU on OVHcloud. With bring your own bucket, the bytes rest wherever your bucket lives while processing stays on EU infrastructure.

Get started

Give your agents storage, extraction, secrets, and search that speak MCP natively. Get started free or read the developer docs. Comparing directly? See Relay vs OneDrive, or read how the same model applies on Google Drive for AI agents.