lt.insight

lt.insight sends natural-language prompts to an LLM that works with the connected MCP servers. It answers queries, builds YAML workflows from a description, refines them from feedback, and describes them. Requires an LLM API key. See MCP.

mcpQuery

const result = await lt.insight.mcpQuery({
  prompt: 'Find all open issues labeled bug',
  tags: ['github'],
  wait: true,
});

Run a natural-language query against the connected MCP servers.

Parameters

Field	Type	Required	Description
`prompt`	`string`	Yes	The query
`tags`	`string[]`	No	Tags that scope which MCP servers are used
`wait`	`boolean`	No	`true` waits for the result; otherwise the call returns once the query starts
`direct`	`boolean`	No	`true` queries the LLM directly, outside workflow orchestration
`context`	`any`	No	Extra context forwarded to the LLM

Returns: LTApiResult<any>

Auth: Optional. The userId is forwarded to the query pipeline when provided.

buildWorkflow

const result = await lt.insight.buildWorkflow({
  prompt: 'Screenshot a URL, analyze the image, and store results in a spreadsheet',
  tags: ['browser', 'vision', 'sheets'],
  wait: true,
});

Generate a YAML workflow from a natural-language description.

Parameters

Field	Type	Required	Description
`prompt`	`string`	Yes	Description of the workflow
`tags`	`string[]`	No	Tags that scope the available MCP tools
`wait`	`boolean`	No	`true` waits for generation to finish
`feedback`	`string`	No	Feedback on a previous generation
`prior_yaml`	`string`	No	YAML from a previous generation
`answers`	`any`	No	Answers to clarifying questions from a prior round
`prior_questions`	`any`	No	Questions from a prior round, for context

Returns: LTApiResult<any>

Auth: Optional. The userId is forwarded to the builder when provided.

refineWorkflow

const result = await lt.insight.refineWorkflow({
  prompt: 'Screenshot and analyze a URL',
  prior_yaml: existingYaml,
  feedback: 'Add a retry step if the screenshot fails',
  wait: true,
});

Refine an existing YAML workflow with feedback.

Parameters

Field	Type	Required	Description
`prompt`	`string`	Yes	Original description
`prior_yaml`	`string`	Yes	The YAML workflow to refine
`feedback`	`string`	Yes	The changes to make
`tags`	`string[]`	No	Tags that scope the available MCP tools
`wait`	`boolean`	No	`true` waits for refinement to finish

Returns: LTApiResult<any>

Auth: Optional. The userId is forwarded to the builder when provided.

describeWorkflow

const result = await lt.insight.describeWorkflow({
  prompt: 'Take a screenshot of a URL and analyze it with a vision model',
});

Generate a readable description and tags for a workflow.

Parameters

Field	Type	Required	Description
`prompt`	`string`	Yes	The workflow prompt or content to describe
`result_title`	`string`	No	Title from the workflow result, for context
`result_summary`	`string`	No	Summary from the workflow result, for context

Returns: LTApiResult<{ description: string, tags: string[] }>

Auth: Not required.