AI Agents

Let AI agents use Syper capabilities safely on behalf of people and businesses.

An AI agent — an assistant, a copilot, an autonomous workflow — can pay, transfer, and read account state for its user. It does so through an Agent Account, never with the user’s own credentials.

Rules for agents

• Own identity. The agent authenticates as itself. Its actions are attributable and revocable.
• Explicit permissions. Grant only the actions the agent needs. See Permissions.
• Spending controls. Every agent that can move money has limits.
• Human confirmation for actions above a threshold, where appropriate.
• Auditability. Everything the agent does appears in activity.

How agents connect

MCP / Agent Tools

For agents running in MCP-compatible environments. Syper capabilities appear as tools the model can call.

SDK

For agents you build yourself in code.

API

For any environment that can make HTTPS requests.

A language model’s output is not an authorisation. Syper enforces the agent’s permissions and limits on every request, regardless of what the model decides.