About Cygrid
Cygrid sells a platform that finds shadow AI, over-scoped tokens and leftover admin rights across SaaS, checks every action by humans, service accounts and AI agents, and secures actions by applying customer policy. Cygrid puts humans, service accounts, API connections and AI agents in one Identity Intelligence Graph. Cygrid connects identity, SaaS, cloud and AI sources into the Identity Intelligence Graph and secures every identity at three enforcement points: before first use, while access exists and at runtime. Cygrid offers one platform to discover, clean up, audit and secure access across SaaS.[1][4][5][6][9]
Cygrid models the action, meaning the permission actually in play, in full context. The graph records what each identity may do, what it actually does and on whose behalf, and policy rules and AI-driven analysis score the risk together. Cygrid discovers the AI agents, MCP servers and connected apps across SaaS, maps each one to the person it acts for, and records who approved what.[4][5][7]
The Cygrid MCP Gateway sits in the tool-call path and audits, allows, rewrites or blocks every agent tool call at runtime, decided by the same graph as every other enforcement point. Every MCP tool call is authenticated through the customer identity provider, tied to the person behind the agent and checked against policy down to the tool and parameter. In the MCP Gateway, credentials are issued just in time, and every call is logged. Cygrid keeps a registry of approved MCP servers with owner, provenance and version, and pauses a server whose tools change until it is re-reviewed. Cygrid verdicts include allow, downscope, block, revoke on approval or grant just in time, and each verdict is written to an immutable decision log. Cygrid issues a scoped credential at the moment of use and lets it expire. Cygrid removes standing admin and grants elevated access on request, for a limited time.[5][8][9][10][11]
Okta, Entra ID or Google Workspace can sync users, groups and offboarding to the MCP Gateway, so access follows the person. Cygrid ingests existing security tools as feeds and does not replace them. Cygrid connectors are read-only by default, with write access only for fixes the customer approves. On approval, Cygrid revokes risky access, orphaned accounts and dormant tokens at the source, blocks unapproved MCP servers and controls which identities AI agents use. Cygrid inventories every OAuth grant, API key, service account and AI app connection, and revokes the risky ones. Cygrid works with ChatGPT, Claude, Claude agents and more, plus any agent that connects through MCP or OAuth.[5][6][8][9][12]
In a finance example, an AI assistant on Drive can read, edit and share files while every delete is blocked. When an agent asks to export 100,000 records, Cygrid blocks the call or reduces it to the 500-record limit before it reaches the app. For e-commerce, Cygrid maps staff, apps and AI agents with access in every Shopify store and flags dormant admins and over-scoped app tokens. Cygrid grants temporary staff, agencies and freelancers time-limited access and removes it automatically when the work ends.[13][14]
Cygrid audits every identity and its access, enforces customer policies on every action, and generates evidence for NIS2, DORA and the EU AI Act from one live graph. Cygrid fills in a register of ICT third parties, including OAuth apps and AI agents, automatically from the graph. Cygrid finds shadow admins whose rights come through groups and never appear in the admin list of the platform, and shows the path from user to group to role. The Cygrid gateway inspects tool calls, not content, and its logs record each action and the person behind it, never prompts. Cygrid is ISO 27001 audited, has SOC 2 Type II in progress and has passed an external penetration test. Cygrid is hosted in Frankfurt across multiple availability zones.[1][4][7][8][12][13][14]
Cygrid on the record