Agent Terms

An agent must act under a valid workspace credential — an sk_live_… API key or an OAuth salty_oat_… token obtained via OAuth 2.1 with PKCE. An agent’s authority is bounded by the single workspace its credential belongs to; it cannot reach other workspaces’ data.

The person or organization that authorizes an agent (the “principal”) is responsible for everything the agent does under their workspace, as if they had done it themselves. Authorize only agents you trust, and scope their access to what they need.

Agents should obtain explicit human consent before high-impact actions — creating an account, upgrading or canceling a plan, raising spend caps, or deleting records. Deliberately, there are no delete_* MCP tools; deletion routes through the REST API or CLI, where a human can review it.

API keys and webhook signing secrets are shown only once. Agents must surface them to the user immediately and must never log, cache, persist, or transmit them to any endpoint other than the Salty API. Treat all credentials as sensitive.

Respect the X-RateLimit-* headers and the Retry-After header on 429 responses, and stay within your plan’s caps. Do not parallelize requests to evade limits. The full declarative policy is published at https://trysalty.com/agent-permissions.json.

Agents should write accurate, relevant CRM data and must not use the workspace to store unlawful, infringing, or abusive content, or to act in ways that violate our Terms of Service or applicable law. Salty validates and signs what it can, but the agent and its principal are responsible for the content written.

Report abuse or suspected credential compromise to abuse@trysalty.com, and security issues per our security policy.