Skip to main content
Harness management is the live inventory of a connected agent: what it is, what it can do, and what it is wired to. If Runtime Events answers “what did my agent do?”, this page answers “what is my agent?”.

​
What you see

  • Identity and status: the agent’s name, description, and connection stage.
  • Prompts: the system prompts and instructions the agent declared, with their configuration.
  • Tools: every toolset and tool schema the agent registered. These become the action catalogue your rules bind to.
  • Connections: the third-party services the agent uses, with credential status only. Rippletide knows which secret names are configured, never their values.
For MCP servers, this page shows the tool inventory, the part that matters for a rules-only subject.

​
Keeping it current

The inventory reflects what the SDK last declared. It re-syncs when the agent starts, or on demand with rippletide connect / the rippletide:sync script. If you add a tool in code, the catalogue updates on the next sync, and rules compiled against the old catalogue will tell you they need attention.

​
Setup evidence vs. runtime proof

A visible inventory proves your agent is configured correctly. It does not prove the agent is running; that is what the heartbeat, tool calls, and live turns in Runtime Events are for. The two pages together tell you both halves of the story.