Keeping your data secure with the AngelList Link
The AngelList MCP lets you query your fund data from AI tools like Claude and ChatGPT. Because that data is sensitive, we built the connection to be read-only, authenticated, and encrypted in transit. This page explains how it works and the steps you can take to keep your access secure.
What is an MCP, in plain terms?
MCP stands for Model Context Protocol — an open standard that lets an AI application securely connect to an outside data source. Think of it as a controlled doorway between your AI client (Claude, ChatGPT, etc.) and your AngelList data.
When you ask your AI client a question about your funds, it requests the relevant records through this doorway, receives them, and forms an answer. The AngelList MCP is read-only: it can look up and return your data, but it cannot create, change, move, or delete anything in your AngelList account.
- Return your funds, firms, vehicles, and capital positions when you ask
- Answer questions using only the data your account is already entitled to
- Operate under your identity, with your permissions
- Write, edit, or delete any AngelList data
- Reach data belonging to other organizations
- Act on its own — every request is driven by you
Security built into the MCP
Authentication, not shared passwords
You connect by signing in with your normal AngelList login through a standard OAuth flow. Your AI client never sees or stores your password — it receives a scoped access token instead, which can be revoked at any time without changing your credentials.
Each client registers itself
Rather than sharing one static key, each AI client registers dynamically and is issued its own credentials for the session. That means access is granted per-client and per-session, so you can identify and revoke an individual connection without disrupting the others.
Encrypted in transit
All traffic between your AI client and AngelList travels over TLS, so data moving through the connection cannot be read in transit — whether you are on the desktop app or Claude Code. Full details on our encryption and data-handling controls are in our Trust Portal.
Least-privilege by design
The MCP runs on the same authentication and authorization infrastructure as the AngelList web app. An authenticated MCP client can only reach the exact data that same user would see in the UI — no more. The entire MCP is read-only, so there is no path to modify records or reach another firm's data.
You sign in the normal way. Connecting uses your standard AngelList login through the OAuth screen — there is no separate key to copy, paste, or store. Your AI client receives a scoped, revocable token, never your password.
Independently audited controls
Our claims are backed by third-party audit, not just our word. The MCP runs on the same audited AngelList infrastructure that already safeguards your fund data.
Certified security controls
AngelList is SOC 2 Type II certified, meaning an independent auditor has verified that our security controls operate effectively over time — not just at a single point in time.
Documentation on request
Our audit report, data-handling practices, subprocessors, and other compliance documentation are available through our Trust Portal. Security reviewers can request access there.
Visit the Trust PortalIf your team needs a signed Data Processing Agreement (DPA), a completed security questionnaire, or a copy of the SOC 2 report for your diligence, reach out to your AngelList contact and we will coordinate through the Trust Portal.
What data the MCP can access
Because the MCP uses the same authentication and authorization infrastructure as the AngelList web app, it can only return data your account is already entitled to see in the UI — nothing more. Your permissions carry over exactly. Typical data it can surface on request includes:
- Funds and vehicles
- Firm and entity details
- Capital positions and commitments
- Investor and LP records you administer
- Portfolio holdings
- Related reporting data held on your behalf
Important: the MCP never pulls data on its own. Nothing is transmitted unless you ask a question that requires it, and only the records relevant to that question are returned. It cannot reach data outside your organization's entitlements, and it has no write access of any kind.
Want tighter scoping for a specific user or use case? Talk to your AngelList contact — we can walk through the access options available to your organization.
Best practices to keep your data secure
Security is shared. AngelList secures the connection; these steps keep your side airtight.
Only ever sign in on the official AngelList screen
Connecting uses your normal AngelList login through OAuth, so enter your credentials only on the genuine AngelList sign-in page. Confirm the URL before authorizing, and never approve a connection request you did not initiate. This protects you against phishing and rogue clients trying to impersonate the flow.
Use a trusted, up-to-date AI client
Connect only from official, current versions of clients you trust — the Claude desktop app, Claude Code, or another reputable MCP client. Download from the official source and keep it updated so you receive the latest security fixes.
Secure the device and account you connect from
Enable full-disk encryption and a screen lock on the machine, and turn on multi-factor authentication for your AngelList login. The MCP acts under your identity, so a secure account and device are your first line of defense.
Understand where your AI client sends data — this is your biggest exposure
Once data leaves AngelList and reaches your AI client, it is governed by that provider's policies, not ours. This is the most important area for you to control, because it sits outside AngelList's boundary. Practical steps:
- Turn off model-training on your inputs. On Claude, consumer chats aren't used for training by default; Team/Enterprise and API/Claude Code data is not trained on. Confirm the setting for your specific plan.
- Prefer a business/enterprise tier for sensitive work — it typically offers stronger data-retention, admin, and audit controls than a personal account.
- Check retention: know how long the provider stores your conversations and whether you can delete them.
- Be mindful of what you paste alongside MCP results, and avoid exporting sensitive answers into less-controlled tools.
Review and revoke access you no longer use
Because access is token-based and per-client, revoking one connection is quick and does not affect the rest. To disconnect:
- In your AI client, open its connector/MCP settings, find the AngelList connection, and remove it — this invalidates its token immediately.
- You can also revoke access from your AngelList account's connected-apps settings, which cuts off all tokens for that connection at once.
- Review connected clients periodically and remove any you no longer use.
Limit who can connect
Grant the AngelList access needed to connect only to the team members who require it, and keep track of who has an active connection. Each person connects under their own login, so the fewer connection points, the smaller the surface to protect.
Frequently asked
Can the MCP change or delete my AngelList data?
No. The MCP is strictly read-only. It can return your data in response to your questions, but it has no ability to create, edit, move, or delete anything in your account.
Does my AI client see my AngelList password?
No. Authentication uses OAuth, so you sign in on AngelList's own screen and the client receives a scoped, revocable access token instead of your password.
Could another firm's data ever reach me — or mine reach them?
No. Access is scoped to your organization only. The connection cannot cross organizational boundaries.
Is the data encrypted?
Yes. All traffic between your AI client and AngelList is encrypted in transit over TLS. Full details of our encryption and data-handling controls are documented in our Trust Portal.
Are you certified or independently audited?
Yes. AngelList is SOC 2 Type II certified. The report and related compliance documentation are available through our Trust Portal, and we can provide a DPA or completed security questionnaire on request.
Could a malicious prompt or tool trick the agent into leaking my data?
The MCP only ever returns data your account is entitled to, and it cannot write or change anything — so it cannot be used to alter your records. As with any AI agent, treat the client itself as a trust boundary: connect only reputable, up-to-date clients, avoid running untrusted tools or instructions alongside it, and be deliberate about where you export results.
What happens to my data once it reaches the AI client?
At that point it is governed by your AI provider's policies, not AngelList's. This is your most important control point: review your provider's data-retention and model-training settings, prefer a business/enterprise tier for sensitive work, and keep inputs out of model training where the option exists.
If something goes wrong, how would I find out?
AngelList maintains incident-response and breach-notification practices as part of our SOC 2 program. If you believe your credentials have been exposed or you notice anything unusual, contact your AngelList representative immediately and we will help you rotate credentials and investigate.
How do I disconnect?
Remove the AngelList connector from your AI client's settings to revoke its token immediately, or revoke it from your AngelList connected-apps settings. Access ends right away.
Still have security questions?
We are happy to walk through anything before you connect, and to provide our SOC 2 report, a DPA, or a completed security questionnaire for your diligence. Reach out to your AngelList contact.
Visit the Trust PortalThis page describes the AngelList MCP and may be updated as the product evolves. It is provided for informational purposes and does not modify any agreement between you and AngelList.