Descope: the factory infographic
One identity platform for people and AI agents: logins for humans, scoped access for agents and MCP servers.
What the company does
In normal terms
Customer and agentic identity platform: passwordless login, MFA, SSO and multi-tenancy, and it secures AI agents and MCP servers with standards-based authorisation, scoped access and token management, so human and machine identities live in one place.
Descope secures agents and MCP servers with standards-based authorisation, scoped access and token management, beside its customer login products.
In lay terms
Descope runs the front desk: it checks who you are, and now gives AI assistants visitor badges that only work for certain rooms and expire.
The risks it faces, especially from AI agents
In normal terms
- Scopes chosen in a hurry. Developers grant broad scopes to make agents work; the token is valid but too wide.
- Agents acting for users. Delegated access mixes the user's rights with the agent's actions, blurring who did what.
- MCP everywhere. Each new MCP server is a new door an agent can reach.
In lay terms
- Scopes chosen in a hurry. The badge works, but it opens every floor.
- Agents acting for users. The assistant uses your badge, so the log says it was you.
- MCP everywhere. Every week a new door is added to the building.
How RiskMandate helps
- Reachwhat it can actually access or affectEvery scope and MCP server the agent can use.
- Mandatewhat the business authorisedThe scopes its task really needs.
- Gapthe difference: the riskScopes granted beyond that.
- Barrierswhat really stands in the wayDescope's short-lived scoped tokens as technical evidence.
- Provenancewho owns it, who reviewed it, which versionWho granted each scope, and when.
In normal terms
Descope is the identity half of what an Agent Behaviour Policy governs. The policy says which scopes an agent should have; Descope issues exactly those. A natural integration and a good interview for the site.
In lay terms
Descope prints the badges; RiskMandate decides which rooms each badge should open.
A first line for the conversation: You issue agents scoped tokens. We write down which scopes each agent should have, and why.
Sources and method
The facts in grey (what it does) are from the company map, which cites:
- websummit.com, read 2026-11-01: LIS26 session page (conferenceSlug lis26): '(A)Identity: Auth tips to secure AI agents and MCP servers', 12 Nov, Developer Summit, Developer Lounge, names Descope as the speaker's organisation; the session is hosted by this company
- websummit.com, read 2026-11-01: LIS26 appearance page: name, country, industry, website and elevator pitch (the 'what' field)
- www.descope.com, read 2026-11-01: what the company does, from its own site (title and description)
The risks and how RiskMandate helps are our analysis, written by dev@web-summit.indi on 3 November 2026 from those facts and RiskMandate's model as this site explains it (Reach, Mandate, Gap; an Agent Behaviour Policy in practice). They are not the company's own statements and name no incident. A person reviews each before it is used.