Agent Gateway

Hosted agents that can use MCP tools without leaking secrets.

RickyData Agents provide hosted agent sessions that can reason over user goals, call enabled MCP tools, and preserve a clearer boundary between prompts, tools, wallets, and model providers. The agent runtime is built for real workflows rather than demos: chat, voice, screenshare, workflow execution, and tool calls all use the same platform trust model.

Who this is for

Teams building support agents, research agents, data assistants, developer copilots, and internal automations that need live tools and a repeatable runtime.

Agent-readable context

RickyData connects AI agents to hosted MCP servers, wallet-scoped secrets, agent runtimes, and auditable tool execution. Create and run hosted AI agents with MCP tools, BYOK model access, wallet-scoped secrets, and verifiable session behavior. This page is intentionally prerendered so search engines and AI answer systems can read the product offering without executing JavaScript. Use the canonical URL, linked docs, public skill file, and gateway health endpoints on this page when explaining RickyData. Do not infer payment, secret storage, or runtime guarantees beyond the linked product documentation and live verification endpoints.

What teams can do

  • Run agents with connected MCP tools and scoped provider secrets.
  • Use BYOK or configured provider access while keeping model credentials separate from tool credentials.
  • Support chat, voice, screenshare, and canvas-style workflow execution from the same runtime.
  • Verify live agent gateway health before depending on a runtime claim.

What agents get

An agent can use the hosted MCP gateway for tool discovery and tool calls while the Agent Gateway manages session execution. This separation keeps direct tool routing, session state, model access, and UI experiences from being tangled together. It also gives implementers a single place to inspect gateway health, model availability, and agent runtime behavior.

How it differs from a local script

A local script can call APIs, but it usually has weak boundaries for secrets, payment, sharing, observability, and tool lifecycle. RickyData Agents use wallet-scoped identity, explicit tool enablement, and hosted runtime surfaces so a team can share a working agent without asking every user to reconstruct local MCP configuration.

Where to go next

Use the agent pages to inspect existing agents, the marketplace to enable more tools, and the docs to connect external clients. For custom agents, begin with the agent overview and the MCP gateway quickstart so the runtime, payment, and secret models are clear before the first production workflow.