- Createcreate_data_room
- Uploadfour paths, one is you
- Organizefolders + Smart Sort
- Inviteheld for your approval
- Askcited answers, engagement
Connect Sifrsys to claude.ai
claude.ai connects to remote tools through custom connectors, which speak the Model Context Protocol (MCP). Sifrsys runs a hosted MCP server at https://sifrsys.com/mcp. There is no API key to paste: claude.ai registers itself with Sifrsys and you sign in through OAuth. Menu names below are from Anthropic's help centre, checked 26 September 2026.
Step 1. Add the connector
On Free, Pro or Max: Customize → Connectors, click +, then Add custom connector. Enter the URL and click Add. Leave Advanced settings empty.
https://sifrsys.com/mcpOn Team or Enterprise, only an Owner can add it: Organization settings → Connectors, Add, hover Custom, choose Web. Members then connect to it individually. On the Free plan Claude allows one custom connector.
Step 2. Sign in and approve
Click Connect. A Sifrsys page opens; sign in (or create a free account) and you see a consent screen naming Claude, listing in plain English what it may do, and asking which organization to grant. Approve it. You can see and revoke the connection later under Settings → Agents & API → Connected apps in Sifrsys.
Step 3. Switch it on in a chat
Open a new chat, click + at the lower left, choose Connectors, and turn Sifrsys on. Claude now sees the 17 Sifrsys tools. If you plan to attach files for Claude to upload, also check that Code execution and file creation is on under Settings → Capabilities (it is on by default on paid Claude plans and off on Free).
Prompts to copy
These are plain requests; Claude picks the tools. The tool each one leads to is named so you can check what happened in the Sifrsys audit log.
Step 4. Create the room
Using Sifrsys, create a data room called "Acme — Seed Round" and give me the room's upload link.Claude gets back the room id and upload_page_url, a link of the form https://sifrsys.com/room/<id>. Keep it; it is the fallback for uploads.
Step 5. Get the documents in
I've attached our deck, financial model and cap table. Upload them to the Acme — Seed Round room.Upload everything in ~/Documents/Acme/Seed to the room.The second prompt will not work in claude.ai, and a well-behaved Claude will say so: it has no access to ~/Documents. Sifrsys tells it what to do instead, which is to send you the room's upload link, ask you to upload the files with the Upload button, and poll get_room_overviewevery 15 to 30 seconds until the room's document count goes up. The next section explains the three routes.
Add this to the room: https://example.com/acme-deck.pdfWrite a one-page data room index listing what each folder contains and upload it to the room as "00 Index.txt".Step 6. Organize, invite, ask
Create folders Corporate, Financials, Legal, Team and Product in the room, then run Smart Sort and show me where it wants to put each file before moving anything.Smart Sort first runs as a preview with a confidence score per document. When you say go ahead, Claude runs it again with apply=true, which moves documents at 0.7 confidence or higher.
Create a group called "Lead investors" with view-only access, watermarking on, downloads off and Q&A on. Invite Jane Doe (jane@fund.example) and Mark Lee (mark@capital.example) into it.Claude creates the group immediately, then gets a pending_approvalback for the invitations. Approve it in Sifrsys under Settings → Agents & API (admins also get an email), then tell Claude:
I approved it. Send the invitations.Give the Lead investors group download access on the Financials folder.Ask the room AI what our monthly net burn was in the last three months of the model, and give me the sheet and cell range it read.Documents become searchable about a minute after upload. The answer comes back with citations to a PDF page or a spreadsheet range, and it is fenced to the documents the connected account may see. How that fence works, and what we measured, is in Can your data room AI see what you cannot?
Who opened the room this week, who hasn't, and is anyone going cold?The anomaly alerts in that report (gone-cold investors, engagement drops) need Room Intelligence, which is on Pro. On other plans the report still returns per-participant engagement.
Files on your laptop: the three ways around it
Everything except the upload is a tool call from Claude to Sifrsys, and tool calls to an MCP server are not subject to Claude's sandbox network rules. Only HTTP requests made by the code Claude runs are. The signed-URL upload is such a request: Claude's code sends the file's bytes to auth.sifrsys.com. Anthropic's documentation lists the default egress setting as package managers only, so that request is blocked until you allow it.
| Criterion | Tool | Limit | Needs outbound network from the agent? |
|---|---|---|---|
| Signed URLAgent PUTs the bytes | request_upload_urls → HTTP PUT → finalize_uploads | 100 MB per file, 25 files per call, URL valid about 5 minutes | Yes, to auth.sifrsys.com |
| InlineContent inside the tool call | upload_inline_document | 4 MB decoded | No |
| From a linkSifrsys fetches it | upload_from_url | 20 MB, https only, no login or cookies | No |
| Upload page hand-offYou upload, the agent waits | upload_page_url, then get_room_overview until document_count rises | Whatever your plan's storage allows | No |
- Allow auth.sifrsys.com and attach the files. On Pro or Max, open Settings → Capabilities and turn on Allow network egress; on Team or Enterprise an Owner sets it under Organization settings → Capabilities, choosing “Allow network egress to package managers and specific domains” and adding
auth.sifrsys.com. Anthropic caps chat attachments at 30 MB per file. This is the only route where Claude moves the bytes itself. - Give Claude links. If the files already sit behind an https link that opens without a login (a pre-signed S3 or GCS link, a Dropbox or Drive direct-download link), Claude calls
upload_from_urland Sifrsys fetches it, up to 20 MB. - Upload them yourself on the room page. Open the
upload_page_urlClaude sent, click Upload, select the files. Claude watches the document count and carries on with organizing and invitations when they arrive. For more than a few files this is usually the fastest route.
Which agent for which situation
| Criterion | claude.ai (web, desktop) | ChatGPT (developer mode) | Claude Code, Cursor, Codex CLI |
|---|---|---|---|
| Sign-in to Sifrsys | OAuth consent screen, no key to paste | OAuth consent screen, no key to paste | API key (sifr_live_…) in a header |
| Files on your own diskThe common case for a seed round | Attach them to the chat and allow auth.sifrsys.com in network settings, or use the upload page | Upload page: you drop the files in, the agent carries on | Works natively: the agent reads the folder and uploads it |
| Files at an https link | upload_from_url, up to 20 MB | upload_from_url, up to 20 MB | Any of the four paths |
| Documents the agent writes (index, FAQ, summary) | upload_inline_document, up to 4 MB | upload_inline_document, up to 4 MB | upload_inline_document, or write the file and upload it |
| Large files | Signed URL (100 MB) only once egress to auth.sifrsys.com is allowed | Upload page | Signed URL, up to 100 MB each |
| Client plans that can connect | Free (one custom connector), Pro, Max, Team, Enterprise | Plus, Pro, Business, Enterprise, Edu, on the web | Any: the client runs on your machine |
| Best for | Founders who live in Claude and have a handful of files | Teams already on ChatGPT; organizing, inviting and asking once files are in | A folder of 30+ files, repeatable setups, anything scripted |
If your documents are a folder on your machine and there are more than a handful of them, a local agent is simpler: see setting up a data room with Claude Code, Cursor or Codex.
What an agent is never allowed to do by default
Every Sifrsys organization has a policy matrix that sets each class of agent action to allow, require approval, or deny. The defaults below are the ones in the server code. An owner or admin can change them under Settings → Agents & API; nothing an agent says in a chat can.
| Criterion | Default | What it covers |
|---|---|---|
| Denied | Refused with denied_by_policy | Deleting or suspending: documents, folders (and merging duplicate folders), participants, groups, and the room itself. Also changing room branding (configure_branding) and publishing the room's participant portal |
| Needs your approval | Returns a pending_approval handle | Sending or resending invitations (invite_participants, resend_invite); widening access: a permission on a specific document or folder (set_permissions), a more permissive group, a later or removed access expiry, reactivating a suspended participant; publishing a Q&A answer to the person who asked, assigning a question, changing its status |
| Allowed | Runs immediately, logged | Reading, creating rooms, uploading (including new versions of a document), renaming and moving documents and folders, removing empty folders, Smart Sort, creating groups, narrowing access, asking the room AI, drafting Q&A answers, restoring a suspended room |
An approval goes to the organization's admins by email and appears under Settings → Agents & API. It is single-use, matched to the exact request the agent made, and expires after 24 hours, so an approval for inviting two people cannot be replayed to invite a third. Every agent action, allowed or not, lands in the room's hash-chained audit log as actor_type=agent with the human who connected it. A key or connection can never do more than the person who created it, and an org-wide switch turns all agent access off at once.
Document text that comes back through ask_room_ai or list_qa is wrapped and labelled as untrusted data, because it was written by the other side of the deal. The threat model explains why.
Agent access is on every Sifrsys plan, the free one included. An agent works inside your plan's limits (the free plan is one room, 100 MB of storage and five participants per room), and the only agent meter is a monthly budget of room-AI questions: 250 on Free, 10,000 on Teams, 50,000 on Pro. Creating, uploading, organizing and inviting are not metered. Bursts are capped at 60 requests a minute per connection. Plans are on the pricing page.
Troubleshooting
The upload fails with a network or 403 error from the sandbox
Claude's code tried to PUT to auth.sifrsys.comand the egress policy refused it. Allow the domain (Settings → Capabilities; on Team/Enterprise, Organization settings → Capabilities) and ask Claude to retry, or ask it to switch to the room's upload page. The upload link it was holding expires after about five minutes, so a retry requests a new one; an upload that was never finalized is cleaned up automatically.
Claude doesn't know about upload_from_url or the upload page
The connector's tool list is read when it connects. Remove Sifrsys under Customize → Connectors (the three-dot menu) and add it again, then start a new chat. Anthropic's documentation gives remove-and-re-add as the way to update a connector.
denied_by_policy
The action is one of the denied classes above. An owner or admin can change the policy under Settings → Agents & API, or do it in the dashboard directly.
limit_reached
The organization is at a plan limit (rooms, participants per room, storage). The free plan allows one room, so a second “create a data room” fails there by design.
The approval never arrives
Approvals expire after 24 hours. Ask Claude to check with get_pending_approvals; if it expired, ask it to make the request again.
Sources
Client steps checked on 26 September 2026 against Anthropic's help centre: Getting started with custom connectors using remote MCP (page dated 11 August 2026) and Create and edit files with Claude (network egress settings, page dated 6 August 2026). Sifrsys tool names, limits and policy defaults are from the MCP server source; the full reference is on the developers page. If Anthropic moves a menu, this page is wrong until we re-check it; the dates are here so you can tell.
The other agent guides: With ChatGPT · With Claude Code, Cursor, Codex · Developer reference · The walkthrough, with the real tool calls