# InTouch base tools — contract reference

> The 44 tools built into InTouch (8.0.3). For each one: what it
> takes, what it refuses to run without, and what it PUBLISHES for a later task to
> reference as `{{taskName.key}}`. Never reference a key that is not listed under
> PUBLISHES — it will resolve to nothing at run time.

Machine-readable: [base-tools.json](./base-tools.json)

## anthropic

Send a prompt to Claude (Anthropic) and use the response.

- **Credential required:** yes (`anthropic`)
- **Publishes:** `inputTokens`, `model`, `outputTokens`, `response`, `stopReason`, `thinking`, `totalTokens`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "topP" : -1.0,
  "topK" : -1,
  "extendedThinking" : false,
  "thinkingBudget" : 10000,
  "stopSequences" : "",
  "outputFile" : "",
  "thinkingOutputFile" : "",
  "timeoutSeconds" : 120
}
```

## anthropic_agent

Autonomous Claude agent that can drive InTouch over multiple steps.

- **Credential required:** yes (`anthropic`)
- **Publishes:** `iterations`, `model`, `response`, `toolCallCount`, `toolHistory`, `totalInputTokens`, `totalOutputTokens`, `totalTokens`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "intouchBaseUrl" : "http://localhost:2200",
  "intouchSessionToken" : "",
  "maxIterations" : 20,
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "extendedThinking" : false,
  "thinkingBudget" : 10000,
  "jobPollIntervalSeconds" : 10,
  "jobMaxWaitSeconds" : 600,
  "outputFile" : "",
  "toolHistoryFile" : "",
  "timeoutSeconds" : 120
}
```

## anthropic_batch

Process many prompts in bulk via Claude's Batch API (cheaper, async).

- **Credential required:** yes (`anthropic`)
- **Publishes:** `batchId`, `errored`, `expired`, `firstResponse`, `processingStatus`, `resultsRetrieved`, `resultsSucceeded`, `succeeded`, `totalRequests`

```json
{
  "inputFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "outputDir" : "",
  "outputFile" : "",
  "pollIntervalSeconds" : 30,
  "maxWaitSeconds" : 3600
}
```

## anthropic_document

Ask Claude about a PDF or document, with source citations.

- **Credential required:** yes (`anthropic`)
- **Publishes:** `citationCount`, `citations`, `inputTokens`, `model`, `outputTokens`, `response`, `stopReason`, `totalTokens`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "documentFiles" : "",
  "imageUrls" : "",
  "enableCitations" : true,
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "extendedThinking" : false,
  "thinkingBudget" : 10000,
  "outputFile" : "",
  "citationsOutputFile" : "",
  "timeoutSeconds" : 300
}
```

## anthropic_tools

Claude with custom tool / function calling.

- **Credential required:** yes (`anthropic`)
- **Publishes:** `iterations`, `model`, `response`, `toolCallCount`, `toolHistory`, `totalInputTokens`, `totalOutputTokens`, `totalTokens`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "toolsFile" : "",
  "toolsJson" : "",
  "maxIterations" : 10,
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "outputFile" : "",
  "toolHistoryFile" : "",
  "timeoutSeconds" : 120
}
```

## archive

Create and open archives — zip, unzip, tar, untar, and list contents without extracting.

- **Publishes:** `count`, `destination`, `entries`, `file`, `fileCount`, `files`, `format`, `operation`, `path`, `size`, `summary`

```json
{
  "operation" : "list",
  "path" : "/tmp/example.zip",
  "destination" : "",
  "include" : ""
}
```

## aws

Run AWS operations via the AWS CLI.

- **Credential required:** yes (`aws`)
- **Publishes:** `commandCount`, `exitCode`, `output`

```json
{
  "service" : "",
  "operation" : "",
  "args" : "",
  "region" : "",
  "outputFormat" : "json",
  "query" : "",
  "outputFile" : "",
  "scriptFile" : "",
  "workingDir" : "",
  "successCodes" : "0",
  "timeoutSeconds" : 300
}
```

## deepseek

Send a prompt to DeepSeek and use the response.

- **Credential required:** yes (`deepseek`)
- **Publishes:** `completionTokens`, `finishReason`, `model`, `promptTokens`, `response`, `totalTokens`, `validationError`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "topP" : -1.0,
  "frequencyPenalty" : 0.0,
  "presencePenalty" : 0.0,
  "responseFormat" : "text",
  "stopSequences" : "",
  "outputFile" : "",
  "timeoutSeconds" : 120,
  "outputSchema" : "",
  "validationRetries" : 0,
  "redactPatterns" : ""
}
```

## discord

Post messages to Discord (outbound).

- **Credential required:** yes (`discord`)
- **Publishes:** nothing — do not reference `{{discordTask.anything}}`

## document_chunker

Split a document into chunks for AI processing.

- **Publishes:** `chunkCount`, `chunks`, `totalChars`, `totalTokenEstimate`

```json
{
  "inputFile" : "",
  "input" : "",
  "splitBy" : "tokens",
  "chunkSize" : 512,
  "chunkOverlap" : 50,
  "separator" : "",
  "metadata" : "",
  "outputFile" : "",
  "maxChunks" : 0
}
```

## email-read

Read messages from an IMAP mailbox (read-only; never marks mail read).

- **Credential required:** yes (`email-read`)
- **Publishes:** `count`, `ids`, `memory`, `messages`, `output`

```json
{
  "name" : "",
  "folder" : "",
  "unreadOnly" : true,
  "sinceDays" : 0,
  "maxMessages" : 25,
  "includeBody" : true,
  "sinceState" : ""
}
```

## embeddings

Turn text into a vector embedding for search and RAG.

- **Credential required:** yes (`openai`)
- **Exactly one of:** `input`, `inputFile`
- **Publishes:** `dimensions`, `embedding`, `embeddingCount`, `model`, `promptTokens`, `totalTokens`

```json
{
  "input" : "",
  "inputFile" : "",
  "model" : "",
  "dimensions" : 0,
  "encodingFormat" : "float",
  "outputFile" : "",
  "timeoutSeconds" : 120
}
```

## fine_tuning

Create and manage model fine-tuning jobs.

- **Credential required:** yes (`openai`)
- **Publishes:** `fineTunedModel`, `workflowId`, `operation`, `response`, `status`

```json
{
  "operation" : "status",
  "model" : "",
  "trainingFile" : "",
  "validationFile" : "",
  "fineTuneJobId" : "",
  "suffix" : "",
  "epochs" : 0,
  "learningRateMultiplier" : 0.0,
  "batchSize" : 0,
  "outputFile" : "",
  "timeoutSeconds" : 300
}
```

## gemini

Send a prompt to Google Gemini and use the response.

- **Credential required:** yes (`gemini`)
- **Publishes:** `candidateTokens`, `finishReason`, `model`, `promptTokens`, `response`, `thinking`, `thinkingTokens`, `totalTokens`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "model" : "",
  "maxOutputTokens" : 4096,
  "temperature" : 1.0,
  "topP" : -1.0,
  "topK" : -1,
  "stopSequences" : "",
  "safetyThreshold" : "BLOCK_MEDIUM_AND_ABOVE",
  "outputFile" : "",
  "thinking" : false,
  "thinkingBudget" : 8192,
  "thinkingOutputFile" : "",
  "timeoutSeconds" : 120
}
```

## google_workspace

Work with Gmail, Calendar, Drive, Sheets, Docs and more.

- **Credential required:** yes (`google_workspace`)
- **Publishes:** `commandCount`, `exitCode`, `output`

```json
{
  "service" : "",
  "operation" : "",
  "args" : "",
  "account" : "",
  "outputFormat" : "json",
  "outputFile" : "",
  "scriptFile" : "",
  "workingDir" : "",
  "successCodes" : "0",
  "timeoutSeconds" : 300,
  "query" : "",
  "max" : 0,
  "messageId" : "",
  "to" : "",
  "subject" : "",
  "bodyHtml" : "",
  "cc" : "",
  "bcc" : "",
  "calendarRange" : "",
  "parent" : "",
  "localPath" : "",
  "fileName" : "",
  "spreadsheetId" : "",
  "range" : "",
  "values" : ""
}
```

## groq

Send a prompt to a Groq-hosted model and use the response.

- **Credential required:** yes (`groq`)
- **Publishes:** `completionTokens`, `finishReason`, `model`, `promptTokens`, `response`, `totalTokens`, `validationError`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "topP" : -1.0,
  "frequencyPenalty" : 0.0,
  "presencePenalty" : 0.0,
  "responseFormat" : "text",
  "stopSequences" : "",
  "outputFile" : "",
  "timeoutSeconds" : 120,
  "outputSchema" : "",
  "validationRetries" : 0,
  "redactPatterns" : ""
}
```

## http

Call any web API or webhook over HTTP (GET, POST, …).

- **Publishes:** `responseBody`, `responseCode`

```json
{
  "url" : "",
  "method" : "GET",
  "headers" : { },
  "body" : "",
  "contentType" : "application/json",
  "timeoutSeconds" : 30,
  "successCodes" : "200,201,202,204",
  "outputFile" : "",
  "followRedirects" : true
}
```

## huggingface

Run a Hugging Face model (summarize, classify, generate, …).

- **Credential required:** yes (`huggingface`)
- **Publishes:** `embeddings`, `entityCount`, `label`, `model`, `rawResponse`, `response`, `score`, `similarities`, `task`, `topToken`

```json
{
  "task" : "text-generation",
  "model" : "",
  "input" : "",
  "inputFile" : "",
  "parameters" : "",
  "candidateLabels" : "",
  "context" : "",
  "waitForModel" : true,
  "outputFile" : "",
  "timeoutSeconds" : 300
}
```

## if

Run steps only when a condition is met.

- **Publishes:** nothing — do not reference `{{ifTask.anything}}`

```json
{
  "arms" : [ ]
}
```

## image_generation

Generate an image from a text prompt.

- **Credential required:** yes (`openai`)
- **Publishes:** `imageCount`, `imageUrl`, `model`, `outputFile`, `revisedPrompt`

```json
{
  "prompt" : "",
  "model" : "",
  "size" : "1024x1024",
  "quality" : "standard",
  "style" : "vivid",
  "n" : 1,
  "responseFormat" : "url",
  "outputFile" : "",
  "timeoutSeconds" : 120
}
```

## iml


- **Credential required:** yes (`iml`)
- **Publishes:** nothing — do not reference `{{imlTask.anything}}`

## intouch-skill

Run an installed InTouch skill.

- **Publishes:** `answer`, `inputTokens`, `model`, `outputTokens`, `skillName`, `toolCallCount`

```json
{
  "skillName" : "",
  "input" : ""
}
```

## intouch_user


- **Credential required:** yes (`intouch_user`)
- **Publishes:** nothing — do not reference `{{intouch_userTask.anything}}`

## line

Send messages via the LINE Messaging API (outbound). Department+.

- **Credential required:** yes (`line`)
- **Publishes:** nothing — do not reference `{{lineTask.anything}}`

## message

Send a notification to your configured channels (email, Slack, Discord, …).

- **Publishes:** nothing — do not reference `{{messageTask.anything}}`

```json
{
  "alertName" : "",
  "subject" : "",
  "body" : "",
  "bodyType" : "text",
  "attachments" : [ ],
  "contactNames" : [ ],
  "userNames" : [ ]
}
```

## mistral

Send a prompt to Mistral and use the response.

- **Credential required:** yes (`mistral`)
- **Publishes:** `completionTokens`, `finishReason`, `model`, `promptTokens`, `response`, `totalTokens`, `validationError`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "topP" : -1.0,
  "frequencyPenalty" : 0.0,
  "presencePenalty" : 0.0,
  "responseFormat" : "text",
  "stopSequences" : "",
  "outputFile" : "",
  "timeoutSeconds" : 120,
  "outputSchema" : "",
  "validationRetries" : 0,
  "redactPatterns" : ""
}
```

## oauth


- **Credential required:** yes (`oauth`)
- **Publishes:** nothing — do not reference `{{oauthTask.anything}}`

## ollama

Send a prompt to a local Ollama model (private, no API cost).

- **Credential required:** yes (`ollama`)
- **Publishes:** `durationMs`, `model`, `outputTokens`, `promptTokens`, `response`, `totalTokens`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "model" : "",
  "temperature" : 0.7,
  "maxTokens" : 4096,
  "topP" : 0.9,
  "outputFile" : "",
  "timeoutSeconds" : 300
}
```

## openai

Send a prompt to OpenAI (GPT) and use the response.

- **Credential required:** yes (`openai`)
- **Publishes:** `completionTokens`, `finishReason`, `model`, `promptTokens`, `response`, `totalTokens`, `validationError`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "topP" : -1.0,
  "frequencyPenalty" : 0.0,
  "presencePenalty" : 0.0,
  "responseFormat" : "text",
  "stopSequences" : "",
  "outputFile" : "",
  "timeoutSeconds" : 120,
  "outputSchema" : "",
  "validationRetries" : 0,
  "redactPatterns" : ""
}
```

## openrouter

Send a prompt to OpenRouter (any vendor's model via one API) and use the response.

- **Credential required:** yes (`openrouter`)
- **Publishes:** `completionTokens`, `finishReason`, `model`, `promptTokens`, `response`, `totalTokens`, `validationError`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "topP" : -1.0,
  "frequencyPenalty" : 0.0,
  "presencePenalty" : 0.0,
  "responseFormat" : "text",
  "stopSequences" : "",
  "outputFile" : "",
  "timeoutSeconds" : 120,
  "outputSchema" : "",
  "validationRetries" : 0,
  "redactPatterns" : ""
}
```

## os

Common OS operations (files, search, disk, date, sleep) — declarative, no shell, fail-loud.

- **Publishes:** `algorithm`, `bytes`, `content`, `count`, `cpuCores`, `date`, `deleted`, `dest`, `disk`, `entries`, `epochMillis`, `exists`, `files`, `free`, `hash`, `java`, `killed`, `lineCount`, `lines`, `matches`, `memory`, `modified`, `os`, `path`, `percentUsed`, `pid`, `replacements`, `set`, `size`, `slept`, `source`, `summary`, `total`, `type`, `used`, `user`, `value`

```json
{
  "operation" : "appendfile",
  "path" : "/tmp/example.log",
  "text" : "hello"
}
```

## runtimeenv

Last resort: run a custom script when no native tool covers the task.

- **Required fields:** `runtimeEnvName`
- **Publishes:** `exitCode`, `output`

```json
{
  "scriptSource" : "inline",
  "script" : "",
  "scriptContent" : "",
  "args" : "",
  "environmentVariables" : "",
  "envVars" : { },
  "credentialNames" : [ ],
  "credentialEnvVarOverrides" : { },
  "successCodes" : "0",
  "timeoutSeconds" : 300,
  "outputFile" : ""
}
```

## slack

Post messages to Slack (outbound).

- **Credential required:** yes (`slack`)
- **Publishes:** nothing — do not reference `{{slackTask.anything}}`

## sleep

Pause the job for a set amount of time.

- **Publishes:** nothing — do not reference `{{sleepTask.anything}}`

```json
{
  "seconds" : 0
}
```

## sms

Send SMS via Telnyx, Twilio, Plivo or AWS SNS (outbound).

- **Credential required:** yes (`sms`)
- **Publishes:** nothing — do not reference `{{smsTask.anything}}`

## smtp

Send email through an SMTP server — the server-wide outbound email transport.

- **Credential required:** yes (`smtp`)
- **Publishes:** nothing — do not reference `{{smtpTask.anything}}`

## speech_to_text

Transcribe an audio file into text.

- **Credential required:** yes (`openai`)
- **Publishes:** `duration`, `language`, `model`, `transcription`

```json
{
  "audioFile" : "",
  "model" : "",
  "language" : "",
  "prompt" : "",
  "responseFormat" : "text",
  "temperature" : 0.0,
  "outputFile" : "",
  "timeoutSeconds" : 300
}
```

## teams

Send messages to Microsoft Teams (outbound). Department+.

- **Credential required:** yes (`teams`)
- **Publishes:** nothing — do not reference `{{teamsTask.anything}}`

## telegram

Send messages via a Telegram bot (outbound).

- **Credential required:** yes (`telegram`)
- **Publishes:** nothing — do not reference `{{telegramTask.anything}}`

## text_to_speech

Convert text into spoken audio.

- **Credential required:** yes (`openai`)
- **Exactly one of:** `input`, `inputFile`
- **Publishes:** `audioSize`, `format`, `inputLength`, `model`, `outputFile`, `voice`

```json
{
  "input" : "",
  "inputFile" : "",
  "model" : "",
  "voice" : "alloy",
  "responseFormat" : "mp3",
  "speed" : 1.0,
  "outputFile" : "",
  "timeoutSeconds" : 120
}
```

## vector_store

Store and search vector embeddings (semantic search).

- **Credential required:** yes (`vector_store`)
- **Publishes:** `operation`, `response`

```json
{
  "operation" : "query",
  "indexName" : "",
  "namespace" : "",
  "queryVector" : "",
  "queryText" : "",
  "topK" : 10,
  "filter" : "",
  "includeMetadata" : true,
  "includeValues" : false,
  "vectors" : "",
  "vectorsFile" : "",
  "deleteIds" : "",
  "deleteFilter" : "",
  "outputFile" : "",
  "timeoutSeconds" : 120
}
```

## whatsapp

Send WhatsApp messages via the Meta Cloud API (outbound). Department+.

- **Credential required:** yes (`whatsapp`)
- **Publishes:** nothing — do not reference `{{whatsappTask.anything}}`

## workflow

Run, enable or disable ANOTHER workflow from inside this one.

- **Required fields:** `ids`
- **Publishes:** nothing — do not reference `{{workflowTask.anything}}`

```json
{
  "operation" : 3,
  "ids" : [ ]
}
```

## xai

Send a prompt to xAI (Grok) and use the response.

- **Credential required:** yes (`xai`)
- **Publishes:** `completionTokens`, `finishReason`, `model`, `promptTokens`, `response`, `totalTokens`, `validationError`

```json
{
  "prompt" : "",
  "promptFile" : "",
  "systemPrompt" : "",
  "systemPromptFile" : "",
  "model" : "",
  "maxTokens" : 4096,
  "temperature" : 1.0,
  "topP" : -1.0,
  "frequencyPenalty" : 0.0,
  "presencePenalty" : 0.0,
  "responseFormat" : "text",
  "stopSequences" : "",
  "outputFile" : "",
  "timeoutSeconds" : 120,
  "outputSchema" : "",
  "validationRetries" : 0,
  "redactPatterns" : ""
}
```