MCP Tools Reference

The Latenode MCP Server exposes tools that AI agents can call to manage scenarios, run executions, and work with connections.

Scenarios

create_scenario

Creates a new scenario with the given name, description, and nodes.

Input

Node object

Output

Notes

• The scenario is always saved with the provided name and description as metadata.
• MCP automatically assigns a display name to every node in the scenario, even if you do not set one explicitly.

---

update_scenario

Replaces an existing scenario by ID. All current nodes are replaced with the ones you provide.

Input

Output

Notes

• Full replacement: the existing node list is discarded and replaced with the one you provide. To make partial changes, first call get_scenario, modify the result, then call update_scenario.
• MCP automatically assigns display names to all nodes, same as in create_scenario.

---

get_scenario

Returns the full definition of a scenario by ID.

Input

Output

---

activate_scenario

Activates or deactivates a scenario.

Input

Output

---

Node types and connections

search_node_types

Searches available node types by keyword. Returns up to 20 results per keyword.

Input

Output: array of node type objects.

---

get_dynamic_node_parameters

Returns additional parameters that become available for a node type once initial parameters are set (e.g. a sheet list after a spreadsheet is selected).

Input

Output: array of additional parameter descriptors (same structure as params in search_node_types).

---

search_connections

Searches for the current user's saved connections by type alias.

Input

Output

---

create_connection

Creates a new connection. For OAuth-based services, returns a URL the user must open to complete authorization manually.

Input

Output

---

Executions

run_scenario_once

Runs a scenario once and returns the execution result.

Input

Output

---

run_node_once

Runs a single node inside a scenario and returns its output. Useful for testing a specific step without running the full scenario.

Input

Output

Notes

• compactOutput defaults to true and returns only the first 100 bytes. Set it to false when you need the full output.
• Use overrideParameters to inject test values without modifying the saved scenario.

---

run_action_node_once

Runs a single action node without a scenario you built in the editor. Pass the node type and parameters; the tool returns the result in the same response. How it behaves for users: Run a single action.

Input

Output

The tool returns the action result in the same call. If the node needs an app connection and none work, the response includes a link to create a new authorization.

Notes

• Distinct from run_node_once: no scenarioId. You do not select a node on the canvas.
• Existing connections are tried first, most recently updated first. If none work, complete the authorization link, then call the tool again.
• Required parameters are validated before the action runs.
• The run is cancelled after 100 seconds.
• The helper scenario is hidden from the Scenarios list. A successful call appears in statistics.

---

get_execution

Returns the status and node outputs of an execution by ID.

Input

Output

Notes

• Use this tool to poll the result after calling run_scenario_once or run_node_once with async: true.

---

get_executions_history

Returns the execution history for a scenario with optional filters.

Input

Output