Access
Identity and authority for people, organizations and AI agents.
(01) Authorize
Access is a complete identity provider with authority built in. It signs in people, services, devices and AI agents over standard protocols, and before an action happens it decides whether that actor, on that authority, may do this exact thing now: allow, deny, or say what is still required. Every decision leaves a signed record that others can check without asking suiss.
(02) The question
“May this actor, on this authority, do this exact thing now?”
(03) Principles
- 01
Two planes, one product
A full identity provider plus explicit, checkable authority. Signing in never grants authority on its own.
- 02
Delegation only narrows
Grants, mandates and delegation chains can never widen. Every action is checked against scope, amount, time, purpose and budget.
- 03
Approvals bound to intent
A human approval is tied to the exact action it approves, so it can be proven later and cannot be reused for something else.
- 04
Revocation that cascades
Revoke, suspend or narrow authority, and the change applies from the moment it is committed, down the whole delegation chain.
- 05
Records you can verify
Signed decision records and an offline verifier let auditors and counterparties check what happened without asking Access.
- 06
Fails closed
If something cannot be verified, the answer is not allow. There is no impersonation, and support never signs in as you.
(04) Capabilities
Identity
- OpenID Connect and OAuth 2.1
- Passkeys, authenticator codes and passwordless email
- Step-up authentication
- Sessions, account lifecycle and recovery
- Hosted pages and embeddable components
Enterprise
- SAML 2.0 and SCIM 2.0
- LDAP, Kerberos, RADIUS and WS-Federation
- OpenID Federation
- Bring your existing identity provider
- Organizations, invitations and verified domains
- Self-service SSO and SCIM setup
Authority
- Grants, mandates and delegation chains
- Allow, deny or require action, with an explanation
- Approvals bound to the exact intent
- Revoke, suspend and narrow with cascade
- Consequence tiers for every action
Proof and audit
- Signed decision receipts
- Portable proof of authority
- Offline verifier
- Tamper-evident, append-only log
- Audit export and replay
- Webhooks, SIEM streams and shared signals
For developers
- Management REST API
- Decision API
- SDKs and an access proxy for your services
- Developer console and decision lab
- Sandbox environment
Governance
- Access reviews
- Separation of duties
- Break-glass access
- Hardware-backed signing keys
- Personal data erasure by crypto-shredding
(05) Agent-native
Take authorization out of the model and put it in a deterministic layer. Agents get identity, bounded authority and a human in the loop when it matters.
- Agent identity with an accountable owner
- Authority budgets and mandates per run
- Sub-agent and agent-to-agent delegation
- MCP authorization server
- Credential vault: the model never sees secrets
- Human approval in the middle of a task
(06) Standards and protocols
- OpenID Connect
- OAuth 2.1
- FAPI 2.0
- PKCE
- DPoP
- SAML 2.0
- SCIM 2.0
- WebAuthn
- LDAP
- Kerberos
- RADIUS
- WS-Federation
- OpenID Federation
- AuthZEN
- CIBA
- Token Exchange
- SSF / CAEP
- SD-JWT
- W3C VC 2.0
- MCP
- A2A
(07) At a glance
- Decisions
- Allow, deny or require action, each with a signed record
- Built with
- Rust, with a shared core for server, browser and mobile
- Keys
- Hardware-backed, kept outside the web processes
- Data
- No physical deletes; personal data erased by crypto-shredding
- SDKs
- TypeScript, Python, Go, Java, .NET and mobile
- Console
- Built to WCAG 2.2 AA
(08) Works with
Events and messagingRelayAccess and Relay run together: every message Access sends goes through Relay.
Personal agentOneHolds One's agent identities, grants and mandates.
Payment infrastructurePayDecides payment authority; Money holds the funds and Pay moves them.
Work coordinationWorkDecides who may approve; Work coordinates the decision.(09) Questions
What is suiss Access?
An identity provider with an authority layer. It signs in people, services and AI agents, and decides before each action whether it may happen, leaving a signed record.
How does Access work with AI agents?
Each agent has its own identity and an accountable owner. Its authority is delegated, can only narrow, and is checked against a budget and purpose on every action. When something needs a person, Access asks for an approval tied to that exact action.
Which protocols does Access support?
OpenID Connect, OAuth 2.1, SAML 2.0, SCIM 2.0, passkeys, LDAP, Kerberos, RADIUS, WS-Federation and MCP authorization, among others.
Is Access available?
Access is in development.