Developer MVP
Policy engine, API, request approvals, mock execution, SDK, dashboard, seeded scenarios and audit evidence.
Autonomy without bounded authority is just uncontrolled access.
AI agents are becoming capable of buying services, invoking paid APIs and moving value. The question is no longer whether they can act. It is whether an organisation can define, enforce and prove what they were allowed to do.
The agent authority problem
Six separate questions. One evidence chain.
Treating these as one vague “payment” problem hides the boundaries that make delegation safe.
01
Authentication
Who is the agent?
02
Authorization
What may the agent do?
03
Intent
What exact action was requested?
04
Policy
Under which constraints?
05
Execution
What actually happened?
06
Audit
Can the organisation prove it?
A precise category
An authority control plane—not another payment rail.
CAPYN starts immediately before execution and ends with evidence of what happened. That boundary keeps the product useful across networks and providers.
CAPYN is not
An unrestricted wallet for agents
A token, DAO or speculative network
A replacement for x402, Solana, Stripe or AP2
A browser-only policy interface
A claim that the v0.1 mock executor is production custody
Financial authority makes every ambiguity visible.
Money forces precise answers about identity, limits, approval, concurrency, replay and audit. Solving those constraints creates a strong base for a more general capability layer.
Exact amounts
Integer minor units and explicit currency.
Explicit mandate
Named capabilities with time-bounded authority.
Constrained spend
Vendor and window limits before execution.
Provable outcome
Decision, approval and execution evidence.
Earn the right to add complexity.
CAPYN will not move into blockchain, token or network complexity until the core authority system is solid and useful to real builders.
Current status
CAPYN v0.1 is a working developer MVP.