Mqttable
Back to docs

MCP is Mqttable’s main interface for agents

The compact MCP model behind Mqttable: 9 external tools, searchable actions, resources, prompts, and Pro workflows.

Jun 24, 2026Mqttable EngineeringMqttable Engineering

Mqttable exposes a compact MCP surface. Agents do not learn dozens of one-off MQTT tools. They learn 9 stable tools, then discover actions, workflows, resources, schemas, examples, and prompts through that surface. This is the main product distinction. The UI is for engineers. MCP is for AI agents that need to do real MQTT work without inventing state or leaking secrets.

Interactive diagram

MCP facade: 9 tools outside, action and workflow catalog inside, resources beside them.

Placeholder for the interactive diagram. Replace with the generated artifact when ready.

placeholder

The 9 MCP tools

  • mqttable_runtime_info reads runtime and MCP status.
  • mqttable_doctor returns passive diagnostics by default.
  • search_mqttable_capabilities finds actions, workflows, resources, and prompts.
  • get_mqttable_schema and get_mqttable_example explain one capability.
  • read_mqttable_resource reads bounded mqttable:// resources.
  • plan_mqttable_action, apply_mqttable_plan, and verify_mqttable_plan handle change workflows.

Actions instead of tool sprawl

Free currently exposes the same compact tool list as Pro and includes 60 Free actions. Pro unlocks additional indexed actions and 7 Pro workflows through the same tools. That means the agent workflow stays stable while product depth grows.

The shortest useful agent path

Call runtime info. Run doctor when discovery is confusing. Search only when the capability id is unknown. Read the schema. Plan the action. Let the human inspect the masked diff. Apply with the confirmation token. Verify by rereading state.